(upbeat music) >> Good morning everybody. Welcome back to Las Vegas. This is day four of theCUBE's wall-to-wall coverage of our Super Bowl, aka AWS re:Invent 2022. I'm here with my co-host, Paul Gillin. My name is Dave Vellante. Sanjay Poonen is in the house, CEO and president of Cohesity. He's sitting in as our guest market watcher, market analyst, you know, deep expertise, new to the job at Cohesity. He was kind enough to sit in, and help us break down what's happening at re:Invent. But Paul, first thing, this morning we heard from Werner Vogels. He was basically given a masterclass on system design. It reminded me of mainframes years ago. When we used to, you know, bury through those IBM blue books and red books. You remember those Sanjay? That's how we- learned back then. >> Oh God, I remember those, Yeah. >> But it made me think, wow, now you know IBM's more of a systems design, nobody talks about IBM anymore. Everybody talks about Amazon. So you wonder, 20 years from now, you know what it's going to be. But >> Well- >> Werner's amazing. >> He pulled out a 24 year old document. >> Yup. >> That he had written early in Amazon's evolution about synchronous design or about essentially distributed architectures that turned out to be prophetic. >> His big thing was nature is asynchronous. So systems are asynchronous. Synchronous is an illusion. It's an abstraction. It's kind of interesting. But, you know- >> Yeah, I mean I've had synonyms for things. Timeless architecture. Werner's an absolute legend. I mean, when you think about folks who've had, you know, impact on technology, you think of people like Jony Ive in design. >> Dave: Yeah. >> You got to think about people like Werner in architecture and just the fact that Andy and the team have been able to keep him engaged that long... I pay attention to his keynote. Peter DeSantis has obviously been very, very influential. And then of course, you know, Adam did a good job, you know, watching from, you know, having watched since I was at the first AWS re:Invent conference, at time was President SAP and there was only a thousand people at this event, okay? Andy had me on stage. I think I was one of the first guest of any tech company in 2011. And to see now this become like, it's a mecca. It's a mother of all IT events, and watch sort of even the transition from Andy to Adam is very special. I got to catch some of Ruba's keynote. So while there's some new people in the mix here, this has become a force of nature. And the last time I was here was 2019, before Covid, watched the last two ones online. But it feels like, I don't know 'about what you guys think, it feels like it's back to 2019 levels. >> I was here in 2019. I feel like this was bigger than 2019 but some people have said that it's about the same. >> I think it was 60,000 versus 50,000. >> Yes. So close. >> It was a little bigger in 2019. But it feels like it's more active. >> And then last year, Sanjay, you weren't here but it was 25,000, which was amazing 'cause it was right in that little space between Omicron, before Omicron hit. But you know, let me ask you a question and this is really more of a question about Amazon's maturity and I know you've been following them since early days. But the way I get the question, number one question I get from people is how is Amazon AWS going to be different under Adam than it was under Andy? What do you think? >> I mean, Adam's not new because he was here before. In some senses he knows the Amazon culture from prior, when he was running sales and marketing prior. But then he took the time off and came back. I mean, this will always be, I think, somewhat Andy's baby, right? Because he was the... I, you know, sent him a text, "You should be really proud of what you accomplished", but you know, I think he also, I asked him when I saw him a few weeks ago "Are you going to come to re:Invent?" And he says, "No, I want to leave this to be Adam's show." And Adam's going to have a slightly different view. His keynotes are probably half the time. It's a little bit more vision. There was a lot more customer stories at the beginning of it. Taking you back to the inspirational pieces of it. I think you're going to see them probably pulling up the stack and not just focused in infrastructure. Many of their platform services are evolved. Many of their, even application services. I'm surprised when I talk to customers. Like Amazon Connect, their sort of call center type technologies, an app layer. It's getting a lot. I mean, I've talked to a couple of Fortune 500 companies that are moving off Ayer to Connect. I mean, it's happening and I did not know that. So it's, you know, I think as they move up the stack, the platform's gotten more... The data centric stack has gotten, and you know, in the area we're working with Cohesity, security, data protection, they're an investor in our company. So this is an important, you know, both... I think tech player and a partner for many companies like us. >> I wonder the, you know, the marketplace... there's been a big push on the marketplace by all the cloud companies last couple of years. Do you see that disrupting the way softwares, enterprise software is sold? >> Oh, for sure. I mean, you have to be a ostrich with your head in the sand to not see this wave happening. I mean, what's it? $150 billion worth of revenue. Even though the growth rates dipped a little bit the last quarter or so, it's still aggregatively between Amazon and Azure and Google, you know, 30% growth. And I think we're still in the second or third inning off a grand 1 trillion or 2 trillion of IT, shifting not all of it to the cloud, but significantly faster. So if you add up all of the big things of the on-premise world, they're, you know, they got to a certain size, their growth is stable, but stalling. These guys are growing significantly faster. And then if you add on top of them, platform companies the data companies, Snowflake, MongoDB, Databricks, you know, Datadog, and then apps companies on top of that. I think the move to the Cloud is inevitable. In SaaS companies, I don't know why you would ever implement a CRM solution on-prem. It's all gone to the Cloud. >> Oh, it is. >> That happened 15 years ago. I mean, begin within three, five years of the advent of Salesforce. And the same thing in HR. Why would you deploy a HR solution now? You've got Workday, you've got, you know, others that are so some of those apps markets are are just never coming back to an on-prem capability. >> Sanjay, I want to ask you, you built a reputation for being able to, you know, forecast accurately, hit your plan, you know, you hit your numbers, you're awesome operator. Even though you have a, you know, technology degree, which you know, that's a two-tool star, multi-tool star. But I call it the slingshot economy. This is like, I mean I've seen probably more downturns than anybody in here, you know, given... Well maybe, maybe- >> Maybe me. >> You and I both. I've never seen anything like this, where where visibility is so unpredictable. The economy is sling-shotting. It's like, oh, hurry up, go Covid, go, go go build, build, build supply, then pull back. And now going forward, now pulling back. Slootman said, you know, on the call, "Hey the guide, is the guide." He said, "we put it out there, We do our best to hit it." But you had CrowdStrike had issues you know, mid-market, ServiceNow. I saw McDermott on the other day on the, on the TV. I just want to pay, you know, buy from the guy. He's so (indistinct) >> But mixed, mixed results, Salesforce, you know, Octa now pre-announcing, hey, they're going to be, or announcing, you know, better visibility, forward guide. Elastic kind of got hit really hard. HPE and Dell actually doing really well in the enterprise. >> Yep. >> 'Course Dell getting killed in the client. But so what are you seeing out there? How, as an executive, do you deal with such poor visibility? >> I think, listen, what the last two or three years have taught us is, you know, with the supply chain crisis, with the surge that people thought you may need of, you know, spending potentially in the pandemic, you have to start off with your tech platform being 10 x better than everybody else. And differentiate, differentiate. 'Cause in a crowded market, but even in a market that's getting tougher, if you're not differentiating constantly through technology innovation, you're going to get left behind. So you named a few places, they're all technology innovators, but even if some of them are having challenges, and then I think you're constantly asking yourselves, how do you move from being a point product to a platform with more and more services where you're getting, you know, many of them moving really fast. In the case of Roe, I like him a lot. He's probably one of the most savvy operators, also that I respect. He calls these speedboats, and you know, his core platform started off with the firewall network security. But he's built now a very credible cloud security, cloud AI security business. And I think that's how you need to be thinking as a tech executive. I mean, if you got core, your core beachhead 10 x better than everybody else. And as you move to adjacencies in these new platforms, have you got now speedboats that are getting to a point where they are competitive advantage? Then as you think of the go-to-market perspective, it really depends on where you are as a company. For a company like our size, we need partners a lot more. Because if we're going to, you know, stand on the shoulders of giants like Isaac Newton said, "I see clearly because I stand on the shoulders giants." I need to really go and cultivate Amazon so they become our lead partner in cloud. And then appropriately Microsoft and Google where I need to. And security. Part of what we announced last week was, last month, yeah, last couple of weeks ago, was the data security alliance with the biggest security players. What was I trying to do with that? First time ever done in my industry was get Palo Alto, CrowdStrike, Wallace, Tenable, CyberArk, Splunk, all to build an alliance with me so I could stand on their shoulders with them helping me. If you're a bigger company, you're constantly asking yourself "how do you make sure you're getting your, like Amazon, their top hundred customers spending more with that?" So I think the the playbook evolves, and I'm watching some of these best companies through this time navigate through this. And I think leadership is going to be tested in enormously interesting ways. >> I'll say. I mean, Snowflake is really interesting because they... 67% growth, which is, I mean, that's best in class for a company that's $2 billion. And, but their guide was still, you know, pretty aggressive. You know, so it's like, do you, you know, when it when it's good times you go, "hey, we can we can guide conservatively and know we can beat it." But when you're not certain, you can't dial down too far 'cause your investors start to bail on you. It's a really tricky- >> But Dave, I think listen, at the end of the day, I mean every CEO should not be worried about the short term up and down in the stock price. You're building a long-term multi-billion dollar company. In the case of Frank, he has, I think I shot to a $10 billion, you know, analytics data warehousing data management company on the back of that platform, because he's eyeing the market that, not just Teradata occupies today, but now Oracle occupies or other databases, right? So his tam as it grows bigger, you're going to have some of these things, but that market's big. I think same with Palo Alto. I mean Datadog's another company, 75% growth. >> Yeah. >> At 20% margins, like almost rule of 95. >> Amazing. >> When they're going after, not just the observability market, they're eating up the sim market, security analytics, the APM market. So I think, you know, that's, you look at these case studies of companies who are going from point product to platforms and are steadily able to grow into new tams. You know, to me that's very inspiring. >> I get it. >> Sanjay: That's what I seek to do at our com. >> I get that it's a marathon, but you know, when you're at VMware, weren't you looking at the stock price every day just out of curiosity? I mean listen, you weren't micromanaging it. >> You do, but at the end of the day, and you certainly look at the days of earnings and so on so forth. >> Yeah. >> Because you want to create shareholder value. >> Yeah. >> I'm not saying that you should not but I think in obsession with that, you know, in a short term, >> Going to kill ya. >> Makes you, you know, sort of myopically focused on what may not be the right thing in the long term. Now in the long arc of time, if you're not creating shareholder value... Look at what happened to Steve Bomber. You needed Satya to come in to change things and he's created a lot of value. >> Dave: Yeah, big time. >> But I think in the short term, my comments were really on the quarter to quarter, but over a four a 12 quarter, if companies are growing and creating profitable growth, they're going to get the valuation they deserve. >> Dave: Yeah. >> Do you the... I want to ask you about something Arvind Krishna said in the previous IBM earnings call, that IT is deflationary and therefore it is resistant to the macroeconomic headwinds. So IT spending should actually thrive in a deflation, in a adverse economic climate. Do you think that's true? >> Not all forms of IT. I pay very close attention to surveys from, whether it's the industry analysts or the Morgan Stanleys, or Goldman Sachs. The financial analysts. And I think there's a gluc in certain sectors that will get pulled back. Traditional view is when the economies are growing people spend on the top line, front office stuff, sales, marketing. If you go and look at just the cloud 100 companies, which are the hottest private companies, and maybe with the public market companies, there's way too many companies focused on sales and marketing. Way too many. I think during a downsizing and recession, that's going to probably shrink some, because they were all built for the 2009 to 2021 era, where it was all about the top line. Okay, maybe there's now a proposition for companies who are focused on cost optimization, supply chain visibility. Security's been intangible, that I think is going to continue to an investment. So I tell, listen, if you are a tech investor or if you're an operator, pay attention to CIO priorities. And right now, in our business at Cohesity, part of the reason we've embraced things like ransomware protection, there is a big focus on security. And you know, by intelligently being a management and a security company around data, I do believe we'll continue to be extremely relevant to CIO budgets. There's a ransomware, 20 ransomware attempts every second. So things of that kind make you relevant in a bank. You have to stay relevant to a buying pattern or else you lose momentum. >> But I think what's happening now is actually IT spending's pretty good. I mean, I track this stuff pretty closely. It's just that expectations were so high and now you're seeing earnings estimates come down and so, okay, and then you, yeah, you've got the, you know the inflationary factors and your discounted cash flows but the market's actually pretty good. >> Yeah. >> You know, relative to other downturns that if this is not a... We're not actually not in a downturn. >> Yeah. >> Not yet anyway. It may be. >> There's a valuation there. >> You have to prepare. >> Not sales. >> Yeah, that's right. >> When I was on CNBC, I said "listen, it's a little bit like that story of Joseph. Seven years of feast, seven years of famine." You have to prepare for potentially your worst. And if it's not the worst, you're in good shape. So will it be a recession 2023? Maybe. You know, high interest rates, inflation, war in Russia, Ukraine, maybe things do get bad. But if you belt tightening, if you're focused in operational excellence, if it's not a recession, you're pleasantly surprised. If it is one, you're prepared for it. >> All right. I'm going to put you in the spot and ask you for predictions. Expert analysis on the World Cup. What do you think? Give us the breakdown. (group laughs) >> As my... I wish India was in the World Cup, but you can't get enough Indians at all to play soccer well enough, but we're not, >> You play cricket, though. >> I'm a US man first. I would love to see one of Brazil, or Argentina. And as a Messi person, I don't know if you'll get that, but it would be really special for Messi to lead, to end his career like Maradonna winning a World Cup. I don't know if that'll happen. I'm probably going to go one of the Latin American countries, if the US doesn't make it far enough. But first loyalty to the US team, and then after one of the Latin American countries. >> And you think one of the Latin American countries is best bet to win or? >> I don't know. It's hard to tell. They're all... What happens now at this stage >> So close, right? >> is anybody could win. >> Yeah. You just have lots of shots of gold. I'm a big soccer fan. It could, I mean, I don't know if the US is favored to win, but if they get far enough, you get to the finals, anybody could win. >> I think they get Netherlands next, right? >> That's tough. >> Really tough. >> But... The European teams are good too, but I would like to see US go far enough, and then I'd like to see Latin America with team one of Argentina, or Brazil. That's my prediction. >> I know you're a big Cricket fan. Are you able to follow Cricket the way you like? >> At god unearthly times the night because they're in Australia, right? >> Oh yeah. >> Yeah. >> I watched the T-20 World Cup, select games of it. Yeah, you know, I'm not rapidly following every single game but the World Cup games, I catch you. >> Yeah, it's good. >> It's good. I mean, I love every sport. American football, soccer. >> That's great. >> You get into basketball now, I mean, I hope the Warriors come back strong. Hey, how about the Warriors Celtics? What do we think? We do it again? >> Well- >> This year. >> I'll tell you what- >> As a Boston Celtics- >> I would love that. I actually still, I have to pay off some folks from Palo Alto office with some bets still. We are seeing unprecedented NBA performance this year. >> Yeah. >> It's amazing. You look at the stats, it's like nothing. I know it's early. Like nothing we've ever seen before. So it's exciting. >> Well, always a pleasure talking to you guys. >> Great to have you on. >> Thanks for having me. >> Thank you. Love the expert analysis. >> Sanjay Poonen. Dave Vellante. Keep it right there. re:Invent 2022, day four. We're winding up in Las Vegas. We'll be right back. You're watching theCUBE, the leader in enterprise and emerging tech coverage. (lighthearted soft music)

>>The cybersecurity landscape continues to be one characterized by a series of point tools designed to do a very specific job, often pretty well, but the mosaic of tooling is grown over the years causing complexity in driving up costs and increasing exposures. So the game of Whackamole continues. Moreover, the way organizations approach security is changing quite dramatically. The cloud, while offering so many advantages, has also created new complexities. The shared responsibility model redefines what the cloud provider secures, for example, the S three bucket and what the customer is responsible for eg properly configuring the bucket. You know, this is all well and good, but because virtually no organization of any size can go all in on a single cloud, that shared responsibility model now spans multiple clouds and with different protocols. Now that of course includes on-prem and edge deployments, making things even more complex. Moreover, the DevOps team is being asked to be the point of execution to implement many aspects of an organization's security strategy. >>This extends to securing the runtime, the platform, and even now containers which can end up anywhere. There's a real need for consolidation in the security industry, and that's part of the answer. We've seen this both in terms of mergers and acquisitions as well as platform plays that cover more and more ground. But the diversity of alternatives and infrastructure implementations continues to boggle the mind with more and more entry points for the attackers. This includes sophisticated supply chain attacks that make it even more difficult to understand how to secure components of a system and how secure those components actually are. The number one challenge CISOs face in today's complex world is lack of talent to address these challenges. And I'm not saying that SecOps pros are not talented, They are. There just aren't enough of them to go around and the adversary is also talented and very creative, and there are more and more of them every day. >>Now, one of the very important roles that a technology vendor can play is to take mundane infrastructure security tasks off the plates of SEC off teams. Specifically we're talking about shifting much of the heavy lifting around securing servers, storage, networking, and other infrastructure and their components onto the technology vendor via r and d and other best practices like supply chain management. And that's what we're here to talk about. Welcome to the second part in our series, A Blueprint for Trusted Infrastructure Made Possible by Dell Technologies and produced by the Cube. My name is Dave Ante and I'm your host now. Previously we looked at what trusted infrastructure means and the role that storage and data protection play in the equation. In this part two of the series, we explore the changing nature of technology infrastructure, how the industry generally in Dell specifically, are adapting to these changes and what is being done to proactively address threats that are increasingly stressing security teams. >>Now today, we continue the discussion and look more deeply into servers networking and hyper-converged infrastructure to better understand the critical aspects of how one company Dell is securing these elements so that dev sec op teams can focus on the myriad new attack vectors and challenges that they faced. First up is Deepak rang Garage Power Edge security product manager at Dell Technologies. And after that we're gonna bring on Mahesh Nagar oim, who was consultant in the networking product management area at Dell. And finally, we're close with Jerome West, who is the product management security lead for HCI hyperconverged infrastructure and converged infrastructure at Dell. Thanks for joining us today. We're thrilled to have you here and hope you enjoy the program. Deepak Arage shoes powered security product manager at Dell Technologies. Deepak, great to have you on the program. Thank you. >>Thank you for having me. >>So we're going through the infrastructure stack and in part one of this series we looked at the landscape overall and how cyber has changed and specifically how Dell thinks about data protection in, in security in a manner that both secures infrastructure and minimizes organizational friction. We also hit on the storage part of the portfolio. So now we want to dig into servers. So my first question is, what are the critical aspects of securing server infrastructure that our audience should be aware of? >>Sure. So if you look at compute in general, right, it has rapidly evolved over the past couple of years, especially with trends toward software defined data centers and with also organizations having to deal with hybrid environments where they have private clouds, public cloud locations, remote offices, and also remote workers. So on top of this, there's also an increase in the complexity of the supply chain itself, right? There are companies who are dealing with hundreds of suppliers as part of their supply chain. So all of this complexity provides a lot of opportunity for attackers because it's expanding the threat surface of what can be attacked, and attacks are becoming more frequent, more severe and more sophisticated. And this has also triggered around in the regulatory and mandates around the security needs. >>And these regulations are not just in the government sector, right? So it extends to critical infrastructure and eventually it also get into the private sector. In addition to this, organizations are also looking at their own internal compliance mandates. And this could be based on the industry in which they're operating in, or it could be their own security postures. And this is the landscape in which servers they're operating today. And given that servers are the foundational blocks of the data center, it becomes extremely important to protect them. And given how complex the modern server platforms are, it's also extremely difficult and it takes a lot of effort. And this means protecting everything from the supply chain to the manufacturing and then eventually the assuring the hardware and software integrity of the platforms and also the operations. And there are very few companies that go to the lens that Dell does in order to secure the server. We truly believe in the notion and the security mentality that, you know, security should enable our customers to go focus on their business and proactively innovate on their business and it should not be a burden to them. And we heavily invest to make that possible for our customers. >>So this is really important because the premise that I set up at the beginning of this was really that I, as of security pro, I'm not a security pro, but if I were, I wouldn't want to be doing all this infrastructure stuff because I now have all these new things I gotta deal with. I want a company like Dell who has the resources to build that security in to deal with the supply chain to ensure the providence, et cetera. So I'm glad you you, you hit on that, but so given what you just said, what does cybersecurity resilience mean from a server perspective? For example, are there specific principles that Dell adheres to that are non-negotiable? Let's say, how does Dell ensure that its customers can trust your server infrastructure? >>Yeah, like when, when it comes to security at Dell, right? It's ingrained in our product, so that's the best way to put it. And security is nonnegotiable, right? It's never an afterthought where we come up with a design and then later on figure out how to go make it secure, right? Our security development life cycle, the products are being designed to counter these threats right from the big. And in addition to that, we are also testing and evaluating these products continuously to identify vulnerabilities. We also have external third party audits which supplement this process. And in addition to this, Dell makes the commitment that we will rapidly respond to any mitigations and vulnerability, any vulnerabilities and exposures found out in the field and provide mitigations and patches for in attacking manner. So this security principle is also built into our server life cycle, right? Every phase of it. >>So we want our products to provide cutting edge capabilities when it comes to security. So as part of that, we are constantly evaluating what our security model is done. We are building on it and continuously improving it. So till a few years ago, our model was primarily based on the N framework of protect, detect and rigor. And it's still aligns really well to that framework, but over the past couple of years, we have seen how computers evolved, how the threads have evolved, and we have also seen the regulatory trends and we recognize the fact that the best security strategy for the modern world is a zero trust approach. And so now when we are building our infrastructure and tools and offerings for customers, first and foremost, they're cyber resilient, right? What we mean by that is they're capable of anticipating threats, withstanding attacks and rapidly recurring from attacks and also adapting to the adverse conditions in which they're deployed. The process of designing these capabilities and identifying these capabilities however, is done through the zero press framework. And that's very important because now we are also anticipating how our customers will end up using these capabilities at there and to enable their own zero trust IT environments and IT zero trusts deployments. We have completely adapted our security approach to make it easier for customers to work with us no matter where they are in their journey towards zero trust option. >>So thank you for that. You mentioned the, this framework, you talked about zero trust. When I think about n I think as well about layered approaches. And when I think about zero trust, I think about if you, if you don't have access to it, you're not getting access, you've gotta earn that, that access and you've got layers and then you still assume that bad guys are gonna get in. So you've gotta detect that and you've gotta response. So server infrastructure security is so fundamental. So my question is, what is Dell providing specifically to, for example, detect anomalies and breaches from unauthorized activity? How do you enable fast and easy or facile recovery from malicious incidents, >>Right? What is that is exactly right, right? Breachers are bound to happen and given how complex our current environment is, it's extremely distributed and extremely connected, right? Data and users are no longer contained with an offices where we can set up a perimeter firewall and say, Yeah, everything within that is good. We can trust everything within it. That's no longer true. The best approach to protect data and infrastructure in the current world is to use a zero trust approach, which uses the principles. Nothing is ever trusted, right? Nothing is trusted implicitly. You're constantly verifying every single user, every single device, and every single access in your system at every single level of your ID environment. And this is the principles that we use on power Edge, right? But with an increased focus on providing granular controls and checks based on the principles of these privileged access. >>So the idea is that service first and foremost need to make sure that the threats never enter and they're rejected at the point of entry, but we recognize breaches are going to occur and if they do, they need to be minimized such that the sphere of damage cost by attacker is minimized so they're not able to move from one part of the network to something else laterally or escalate their privileges and cause more damage, right? So the impact radius for instance, has to be radius. And this is done through features like automated detection capabilities and automation, automated remediation capabilities. So some examples are as part of our end to end boot resilience process, we have what they call a system lockdown, right? We can lock down the configuration of the system and lock on the form versions and all changes to the system. And we have capabilities which automatically detect any drift from that lockdown configuration and we can figure out if the drift was caused to authorized changes or unauthorized changes. >>And if it is an unauthorize change can log it, generate security alerts, and we even have capabilities to automatically roll the firm where, and always versions back to a known good version and also the configurations, right? And this becomes extremely important because as part of zero trust, we need to respond to these things at machine speed and we cannot do it at a human speed. And having these automated capabilities is a big deal when achieving that zero trust strategy. And in addition to this, we also have chassis inclusion detection where if the chassis, the box, the several box is opened up, it logs alerts, and you can figure out even later if there's an AC power cycle, you can go look at the logs to see that the box is opened up and figure out if there was a, like a known authorized access or some malicious actor opening and chain something in your system. >>Great, thank you for that lot. Lot of detail and and appreciate that. I want to go somewhere else now cuz Dell has a renowned supply chain reputation. So what about securing the, the supply chain and the server bill of materials? What does Dell specifically do to track the providence of components it uses in its systems so that when the systems arrive, a customer can be a hundred percent certain that that system hasn't been compromised, >>Right? And we've talked about how complex the modern supply chain is, right? And that's no different for service. We have hundreds of confidence on the server and a lot of these form where in order to be configured and run and this former competence could be coming from third parties suppliers. So now the complexity that we are dealing with like was the end to end approach and that's where Dell pays a lot of attention into assuring the security approach approaching and it starts all the way from sourcing competence, right? And then through the design and then even the manufacturing process where we are wetting the personnel leather factories and wetting the factories itself. And the factories also have physical controls, physical security controls built into them and even shipping, right? We have GPS tagging of packages. So all of this is built to ensure supply chain security. >>But a critical aspect of this is also making sure that the systems which are built in the factories are delivered to the customers without any changes or any tapper. And we have a feature called the secure component verification, which is capable of doing this. What the feature does this, when the system gets built in a factory, it generates an inventory of all the competence in the system and it creates a cryptographic certificate based on the signatures presented to this by the competence. And this certificate is stored separately and sent to the customers separately from the system itself. So once the customers receive the system at their end, they can run out to, it generates an inventory of the competence on the system at their end and then compare it to the golden certificate to make sure nothing was changed. And if any changes are detected, we can figure out if there's an authorized change or unauthorize change. >>Again, authorized changes could be like, you know, upgrades to the drives or memory and ized changes could be any sort of temper. So that's the supply chain aspect of it and bill of metal use is also an important aspect to galing security, right? And we provide a software bill of materials, which is basically a list of ingredients of all the software pieces in the platform. So what it allows our customers to do is quickly take a look at all the different pieces and compare it to the vulnerability database and see if any of the vulner which have been discovered out in the wild affected platform. So that's a quick way of figuring out if the platform has any known vulnerabilities and it has not been patched. >>Excellent. That's really good. My last question is, I wonder if you, you know, give us the sort of summary from your perspective, what are the key strengths of Dell server portfolio from a security standpoint? I'm really interested in, you know, the uniqueness and the strong suit that Dell brings to the table, >>Right? Yeah. We have talked enough about the complexity of the environment and how zero risk is necessary for the modern ID environment, right? And this is integral to Dell powered service. And as part of that like you know, security starts with the supply chain. We already talked about the second component verification, which is a beneath feature that Dell platforms have. And on top of it we also have a silicon place platform mode of trust. So this is a key which is programmed into the silicon on the black service during manufacturing and can never be changed after. And this immutable key is what forms the anchor for creating the chain of trust that is used to verify everything in the platform from the hardware and software integrity to the boot, all pieces of it, right? In addition to that, we also have a host of data protection features. >>Whether it is protecting data at risk in news or inflight, we have self encrypting drives which provides scalable and flexible encryption options. And this couple with external key management provides really good protection for your data address. External key management is important because you know, somebody could physically steam the server walk away, but then the keys are not stored on the server, it stood separately. So that provides your action layer of security. And we also have dual layer encryption where you can compliment the hardware encryption on the secure encrypted drives with software level encryption. Inion to this we have identity and access management features like multifactor authentication, single sign on roles, scope and time based access controls, all of which are critical to enable that granular control and checks for zero trust approach. So I would say like, you know, if you look at the Dell feature set, it's pretty comprehensive and we also have the flexibility built in to meet the needs of all customers no matter where they fall in the spectrum of, you know, risk tolerance and security sensitivity. And we also have the capabilities to meet all the regulatory requirements and compliance requirements. So in a nutshell, I would say that you know, Dell Power Service cyber resident infrastructure helps accelerate zero tested option for customers. >>Got it. So you've really thought this through all the various things that that you would do to sort of make sure that your server infrastructure is secure, not compromised, that your supply chain is secure so that your customers can focus on some of the other things that they have to worry about, which are numerous. Thanks Deepak, appreciate you coming on the cube and participating in the program. >>Thank you for having >>You're welcome. In a moment I'll be back to dig into the networking portion of the infrastructure. Stay with us for more coverage of a blueprint for trusted infrastructure and collaboration with Dell Technologies on the cube, your leader in enterprise and emerging tech coverage. We're back with a blueprint for trusted infrastructure and partnership with Dell Technologies in the cube. And we're here with Mahesh Nager, who is a consultant in the area of networking product management at Dell Technologies. Mahesh, welcome, good to see you. >>Hey, good morning Dell's, nice to meet, meet to you as well. >>Hey, so we've been digging into all the parts of the infrastructure stack and now we're gonna look at the all important networking components. Mahesh, when we think about networking in today's environment, we think about the core data center and we're connecting out to various locations including the cloud and both the near and the far edge. So the question is from Dell's perspective, what's unique and challenging about securing network infrastructure that we should know about? >>Yeah, so few years ago IT security and an enterprise was primarily putting a wrapper around data center out because it was constrained to an infrastructure owned and operated by the enterprise for the most part. So putting a rapid around it like a parameter or a firewall was a sufficient response because you could basically control the environment and data small enough control today with the distributed data, intelligent software, different systems, multi-cloud environment and asset service delivery, you know, the infrastructure for the modern era changes the way to secure the network infrastructure In today's, you know, data driven world, it operates everywhere and data has created and accessed everywhere so far from, you know, the centralized monolithic data centers of the past. The biggest challenge is how do we build the network infrastructure of the modern era that are intelligent with automation enabling maximum flexibility and business agility without any compromise on the security. We believe that in this data era, the security transformation must accompany digital transformation. >>Yeah, that's very good. You talked about a couple of things there. Data by its very nature is distributed. There is no perimeter anymore, so you can't just, as you say, put a rapper around it. I like the way you phrase that. So when you think about cyber security resilience from a networking perspective, how do you define that? In other words, what are the basic principles that you adhere to when thinking about securing network infrastructure for your customers? >>So our belief is that cybersecurity and cybersecurity resilience, they need to be holistic, they need to be integrated, scalable, one that span the entire enterprise and with a co and objective and policy implementation. So cybersecurity needs to span across all the devices and running across any application, whether the application resets on the cloud or anywhere else in the infrastructure. From a networking standpoint, what does it mean? It's again, the same principles, right? You know, in order to prevent the threat actors from accessing changing best destroy or stealing sensitive data, this definition holds good for networking as well. So if you look at it from a networking perspective, it's the ability to protect from and withstand attacks on the networking systems as we continue to evolve. This will also include the ability to adapt and recover from these attacks, which is what cyber resilience aspect is all about. So cybersecurity best practices, as you know, is continuously changing the landscape primarily because the cyber threats also continue to evolve. >>Yeah, got it. So I like that. So it's gotta be integrated, it's gotta be scalable, it's gotta be comprehensive, comprehensive and adaptable. You're saying it can't be static, >>Right? Right. So I think, you know, you had a second part of a question, you know, that says what do we, you know, what are the basic principles? You know, when you think about securing network infrastructure, when you're looking at securing the network infrastructure, it revolves around core security capability of the devices that form the network. And what are these security capabilities? These are access control, software integrity and vulnerability response. When you look at access control, it's to ensure that only the authenticated users are able to access the platform and they're able to access only the kind of the assets that they're authorized to based on their user level. Now accessing a network platform like a switch or a rotor for example, is typically used for say, configuration and management of the networking switch. So user access is based on say roles for that matter in a role based access control, whether you are a security admin or a network admin or a storage admin. >>And it's imperative that logging is enable because any of the change to the configuration is actually logged and monitored as that. Talking about software's integrity, it's the ability to ensure that the software that's running on the system has not been compromised. And, and you know, this is important because it could actually, you know, get hold of the system and you know, you could get UND desire results in terms of say validation of the images. It's, it needs to be done through say digital signature. So, so it's important that when you're talking about say, software integrity, a, you are ensuring that the platform is not compromised, you know, is not compromised and be that any upgrades, you know, that happens to the platform is happening through say validated signature. >>Okay. And now, now you've now, so there's access control, software integrity, and I think you, you've got a third element which is i I think response, but please continue. >>Yeah, so you know, the third one is about civil notability. So we follow the same process that's been followed by the rest of the products within the Dell product family. That's to report or identify, you know, any kind of a vulnerability that's being addressed by the Dell product security incident response team. So the networking portfolio is no different, you know, it follows the same process for identification for tri and for resolution of these vulnerabilities. And these are addressed either through patches or through new reasons via networking software. >>Yeah, got it. Okay. So I mean, you didn't say zero trust, but when you were talking about access control, you're really talking about access to only those assets that people are authorized to access. I know zero trust sometimes is a buzzword, but, but you I think gave it, you know, some clarity there. Software integrity, it's about assurance validation, your digital signature you mentioned and, and that there's been no compromise. And then how you respond to incidents in a standard way that can fit into a security framework. So outstanding description, thank you for that. But then the next question is, how does Dell networking fit into the construct of what we've been talking about Dell trusted infrastructure? >>Okay, so networking is the key element in the Dell trusted infrastructure. It provides the interconnect between the service and the storage world. And you know, it's part of any data center configuration for a trusted infrastructure. The network needs to have access control in place where only the authorized nels are able to make change to the network configuration and logging off any of those changes is also done through the logging capabilities. Additionally, we should also ensure that the configuration should provide network isolation between say the management network and the data traffic network because they need to be separate and distinct from each other. And furthermore, even if you look at the data traffic network and now you have things like segmentation isolated segments and via VRF or, or some micro segmentation via partners, this allows various level of security for each of those segments. So it's important you know, that, that the network infrastructure has the ability, you know, to provide all this, this services from a Dell networking security perspective, right? >>You know, there are multiple layer of defense, you know, both at the edge and in the network in this hardware and in the software and essentially, you know, a set of rules and a configuration that's designed to sort of protect the integrity, confidentiality, and accessibility of the network assets. So each network security layer, it implements policies and controls as I said, you know, including send network segmentation. We do have capabilities sources, centralized management automation and capability and scalability for that matter. Now you add all of these things, you know, with the open networking standards or software, different principles and you essentially, you know, reach to the point where you know, you're looking at zero trust network access, which is essentially sort of a building block for increased cloud adoption. If you look at say that you know the different pillars of a zero trust architecture, you know, if you look at the device aspect, you know, we do have support for security for example, we do have say trust platform in a trusted platform models tpms on certain offer products and you know, the physical security know plain, simple old one love port enable from a user trust perspective, we know it's all done via access control days via role based access control and say capability in order to provide say remote authentication or things like say sticky Mac or Mac learning limit and so on. >>If you look at say a transport and decision trust layer, these are essentially, you know, how do you access, you know, this switch, you know, is it by plain hotel net or is it like secure ssh, right? And you know, when a host communicates, you know, to the switch, we do have things like self-signed or is certificate authority based certification. And one of the important aspect is, you know, in terms of, you know, the routing protocol, the routing protocol, say for example BGP for example, we do have the capability to support MD five authentication between the b g peers so that there is no, you know, manages attack, you know, to the network where the routing table is compromised. And the other aspect is about second control plane is here, you know, you know, it's, it's typical that if you don't have a control plane here, you know, it could be flooded and you know, you know, the switch could be compromised by city denial service attacks. >>From an application test perspective, as I mentioned, you know, we do have, you know, the application specific security rules where you could actually define, you know, the specific security rules based on the specific applications, you know, that are running within the system. And I did talk about, say the digital signature and the cryptographic check that we do for authentication and for, I mean rather for the authenticity and the validation of, you know, of the image and the BS and so on and so forth. Finally, you know, the data trust, we are looking at, you know, the network separation, you know, the network separation could happen or VRF plain old wheel Ls, you know, which can bring about sales multi 10 aspects. We talk about some microsegmentation as it applies to nsx for example. The other aspect is, you know, we do have, with our own smart fabric services that's enabled in a fabric, we have a concept of c cluster security. So all of this, you know, the different pillars, they sort of make up for the zero trust infrastructure for the networking assets of an infrastructure. >>Yeah. So thank you for that. There's a, there's a lot to unpack there. You know, one of the premise, the premise really of this, this, this, this segment that we're setting up in this series is really that everything you just mentioned, or a lot of things you just mentioned used to be the responsibility of the security team. And, and the premise that we're putting forth is that because security teams are so stretched thin, you, you gotta shift the vendor community. Dell specifically is shifting a lot of those tasks to their own r and d and taking care of a lot of that. So, cuz scop teams got a lot of other stuff to, to worry about. So my question relates to things like automation, which can help and scalability, what about those topics as it relates to networking infrastructure? >>Okay, our >>Portfolio, it enables state of the automation software, you know, that enables simplifying of the design. So for example, we do have, you know, you know the fabric design center, you know, a tool that automates the design of the fabric and you know, from a deployment and you know, the management of the network infrastructure that are simplicities, you know, using like Ansible s for Sonic for example are, you know, for a better sit and tell story. You know, we do have smart fabric services that can automate the entire fabric, you know, for a storage solution or for, you know, for one of the workloads for example. Now we do help reduce the complexity by closely integrating the management of the physical and the virtual networking infrastructure. And again, you know, we have those capabilities using Sonic or Smart Traffic services. If you look at Sonic for example, right? >>It delivers automated intent based secure containerized network and it has the ability to provide some network visibility and Avan has and, and all of these things are actually valid, you know, for a modern networking infrastructure. So now if you look at Sonic, you know, it's, you know, the usage of those tools, you know, that are available, you know, within the Sonic no is not restricted, you know, just to the data center infrastructure is, it's a unified no, you know, that's well applicable beyond the data center, you know, right up to the edge. Now if you look at our north from a smart traffic OS 10 perspective, you know, as I mentioned, we do have smart traffic services which essentially, you know, simplifies the deployment day zero, I mean rather day one, day two deployment expansion plans and the lifecycle management of our conversion infrastructure and hyper and hyper conversion infrastructure solutions. And finally, in order to enable say, zero touch deployment, we do have, you know, a VP solution with our SD van capability. So these are, you know, ways by which we bring down the complexity by, you know, enhancing the automation capability using, you know, a singular loss that can expand from a data center now right to the edge. >>Great, thank you for that. Last question real quick, just pitch me, what can you summarize from your point of view, what's the strength of the Dell networking portfolio? >>Okay, so from a Dell networking portfolio, we support capabilities at multiple layers. As I mentioned, we're talking about the physical security for examples, say disabling of the unused interface. Sticky Mac and trusted platform modules are the things that to go after. And when you're talking about say secure boot for example, it delivers the authenticity and the integrity of the OS 10 images at the startup. And Secure Boot also protects the startup configuration so that, you know, the startup configuration file is not compromised. And Secure port also enables the workload of prediction, for example, that is at another aspect of software image integrity validation, you know, wherein the image is data for the digital signature, you know, prior to any upgrade process. And if you are looking at secure access control, we do have things like role based access control, SSH to the switches, control plane access control that pre do tags and say access control from multifactor authentication. >>We do have various tech ads for entry control to the network and things like CSE and PRV support, you know, from a federal perspective we do have say logging wherein, you know, any event, any auditing capabilities can be possible by say looking at the clog service, you know, which are pretty much in our transmitter from the devices overts for example, and last we talked about say network segment, you know, say network separation and you know, these, you know, separation, you know, ensures that are, that is, you know, a contained say segment, you know, for a specific purpose or for the specific zone and, you know, just can be implemented by a, a micro segmentation, you know, just a plain old wheel or using virtual route of framework VR for example. >>A lot there. I mean I think frankly, you know, my takeaway is you guys do the heavy lifting in a very complicated topic. So thank you so much for, for coming on the cube and explaining that in in quite some depth. Really appreciate it. >>Thank you indeed. >>Oh, you're very welcome. Okay, in a moment I'll be back to dig into the hyper-converged infrastructure part of the portfolio and look at how when you enter the world of software defined where you're controlling servers and storage and networks via software led system, you could be sure that your infrastructure is trusted and secure. You're watching a blueprint for trusted infrastructure made possible by Dell Technologies and collaboration with the cube, your leader in enterprise and emerging tech coverage, your own west product management security lead at for HCI at Dell Technologies hyper-converged infrastructure. Jerome, welcome. >>Thank you Dave. >>Hey Jerome, in this series of blueprint for trusted infrastructure, we've been digging into the different parts of the infrastructure stack, including storage servers and networking, and now we want to cover hyperconverged infrastructure. So my first question is, what's unique about HCI that presents specific security challenges? What do we need to know? >>So what's unique about hyper-converge infrastructure is the breadth of the security challenge. We can't simply focus on a single type of IT system. So like a server or storage system or a virtualization piece of software, software. I mean HCI is all of those things. So luckily we have excellent partners like VMware, Microsoft, and internal partners like the Dell Power Edge team, the Dell storage team, the Dell networking team, and on and on. These partnerships in these collaborations are what make us successful from a security standpoint. So let me give you an example to illustrate. In the recent past we're seeing growing scope and sophistication in supply chain attacks. This mean an attacker is going to attack your software supply chain upstream so that hopefully a piece of code, malicious code that wasn't identified early in the software supply chain is distributed like a large player, like a VMware or Microsoft or a Dell. So to confront this kind of sophisticated hard to defeat problem, we need short term solutions and we need long term solutions as well. >>So for the short term solution, the obvious thing to do is to patch the vulnerability. The complexity is for our HCI portfolio. We build our software on VMware, so we would have to consume a patch that VMware would produce and provide it to our customers in a timely manner. Luckily VX rail's engineering team has co engineered a release process with VMware that significantly shortens our development life cycle so that VMware would produce a patch and within 14 days we will integrate our own code with the VMware release we will have tested and validated the update and we will give an update to our customers within 14 days of that VMware release. That as a result of this kind of rapid development process, VHA had over 40 releases of software updates last year for a longer term solution. We're partnering with VMware and others to develop a software bill of materials. We work with VMware to consume their software manifest, including their upstream vendors and their open source providers to have a comprehensive list of software components. Then we aren't caught off guard by an unforeseen vulnerability and we're more able to easily detect where the software problem lies so that we can quickly address it. So these are the kind of relationships and solutions that we can co engineer with effective collaborations with our, with our partners. >>Great, thank you for that. That description. So if I had to define what cybersecurity resilience means to HCI or converged infrastructure, and to me my takeaway was you gotta have a short term instant patch solution and then you gotta do an integration in a very short time, you know, two weeks to then have that integration done. And then longer term you have to have a software bill of materials so that you can ensure the providence of all the components help us. Is that a right way to think about cybersecurity resilience? Do you have, you know, a additives to that definition? >>I do. I really think that's site cybersecurity and resilience for hci because like I said, it has sort of unprecedented breadth across our portfolio. It's not a single thing, it's a bit of everything. So really the strength or the secret sauce is to combine all the solutions that our partner develops while integrating them with our own layer. So let me, let me give you an example. So hci, it's a, basically taking a software abstraction of hardware functionality and implementing it into something called the virtualized layer. It's basically the virtual virtualizing hardware functionality, like say a storage controller, you could implement it in hardware, but for hci, for example, in our VX rail portfolio, we, our Vxl product, we integrated it into a product called vsan, which is provided by our partner VMware. So that portfolio of strength is still, you know, through our, through our partnerships. >>So what we do, we integrate these, these security functionality and features in into our product. So our partnership grows to our ecosystem through products like VMware, products like nsx, Horizon, Carbon Black and vSphere. All of them integrate seamlessly with VMware and we also leverage VMware's software, part software partnerships on top of that. So for example, VX supports multifactor authentication through vSphere integration with something called Active Directory Federation services for adfs. So there's a lot of providers that support adfs including Microsoft Azure. So now we can support a wide array of identity providers such as Off Zero or I mentioned Azure or Active Directory through that partnership. So we can leverage all of our partners partnerships as well. So there's sort of a second layer. So being able to secure all of that, that provides a lot of options and flexibility for our customers. So basically to summarize my my answer, we consume all of the security advantages of our partners, but we also expand on them to make a product that is comprehensively secured at multiple layers from the hardware layer that's provided by Dell through Power Edge to the hyper-converged software that we build ourselves to the virtualization layer that we get through our partnerships with Microsoft and VMware. >>Great, I mean that's super helpful. You've mentioned nsx, Horizon, Carbon Black, all the, you know, the VMware component OTH zero, which the developers are gonna love. You got Azure identity, so it's really an ecosystem. So you may have actually answered my next question, but I'm gonna ask it anyway cuz you've got this software defined environment and you're managing servers and networking and storage with this software led approach, how do you ensure that the entire system is secure end to end? >>That's a really great question. So the, the answer is we do testing and validation as part of the engineering process. It's not just bolted on at the end. So when we do, for example, VxRail is the market's only co engineered solution with VMware, other vendors sell VMware as a hyper converged solution, but we actually include security as part of the co-engineering process with VMware. So it's considered when VMware builds their code and their process dovetails with ours because we have a secure development life cycle, which other products might talk about in their discussions with you that we integrate into our engineering life cycle. So because we follow the same framework, all of the, all of the codes should interoperate from a security standpoint. And so when we do our final validation testing when we do a software release, we're already halfway there in ensuring that all these features will give the customers what we promised. >>That's great. All right, let's, let's close pitch me, what would you say is the strong suit summarize the, the strengths of the Dell hyper-converged infrastructure and converged infrastructure portfolio specifically from a security perspective? Jerome? >>So I talked about how hyper hyper-converged infrastructure simplifies security management because basically you're gonna take all of these features that are abstracted in in hardware, they're now abstracted in the virtualization layer. Now you can manage them from a single point of view, whether it would be, say, you know, in for VX rail would be b be center, for example. So by abstracting all this, you make it very easy to manage security and highly flexible because now you don't have limitations around a single vendor. You have a multiple array of choices and partnerships to select. So I would say that is the, the key to making it to hci. Now, what makes Dell the market leader in HCI is not only do we have that functionality, but we also make it exceptionally useful to you because it's co engineered, it's not bolted on. So I gave the example of spo, I gave the example of how we, we modify our software release process with VMware to make it very responsive. >>A couple of other features that we have specific just to HCI are digitally signed LCM updates. This is an example of a feature that we have that's only exclusive to Dell that's not done through a partnership. So we digitally signed our software updates so the user can be sure that the, the update that they're installing into their system is an authentic and unmodified product. So we give it a Dell signature that's invalidated prior to installation. So not only do we consume the features that others develop in a seamless and fully validated way, but we also bolt on our own a specific HCI security features that work with all the other partnerships and give the user an exceptional security experience. So for, for example, the benefit to the customer is you don't have to create a complicated security framework that's hard for your users to use and it's hard for your system administrators to manage it all comes in a package. So it, it can be all managed through vCenter, for example, or, and then the specific hyper, hyper-converged functions can be managed through VxRail manager or through STDC manager. So there's very few pains of glass that the, the administrator or user ever has to worry about. It's all self contained and manageable. >>That makes a lot of sense. So you've got your own infrastructure, you're applying your best practices to that, like the digital signatures, you've got your ecosystem, you're doing co-engineering with the ecosystems, delivering security in a package, minimizing the complexity at the infrastructure level. The reason Jerome, this is so important is because SecOps teams, you know, they gotta deal with cloud security, they gotta deal with multiple clouds. Now they have their shared responsibility model going across multiple cl. They got all this other stuff that they have to worry, they gotta secure the containers and the run time and and, and, and, and the platform and so forth. So they're being asked to do other things. If they have to worry about all the things that you just mentioned, they'll never get, you know, the, the securities is gonna get worse. So what my takeaway is, you're removing that infrastructure piece and saying, Okay guys, you now can focus on those other things that is not necessarily Dell's, you know, domain, but you, you know, you can work with other partners to and your own teams to really nail that. Is that a fair summary? >>I think that is a fair summary because absolutely the worst thing you can do from a security perspective is provide a feature that's so unusable that the administrator disables it or other key security features. So when I work with my partners to define, to define and develop a new security feature, the thing I keep foremost in mind is, will this be something our users want to use and our administrators want to administer? Because if it's not, if it's something that's too difficult or onerous or complex, then I try to find ways to make it more user friendly and practical. And this is a challenge sometimes because we are, our products operate in highly regulated environments and sometimes they have to have certain rules and certain configurations that aren't the most user friendly or management friendly. So I, I put a lot of effort into thinking about how can we make this feature useful while still complying with all the regulations that we have to comply with. And by the way, we're very successful in a highly regulated space. We sell a lot of VxRail, for example, into the Department of Defense and banks and, and other highly regulated environments and we're very successful there. >>Excellent. Okay, Jerome, thanks. We're gonna leave it there for now. I'd love to have you back to talk about the progress that you're making down the road. Things always, you know, advance in the tech industry and so would appreciate that. >>I would look forward to it. Thank you very much, Dave. >>You're really welcome. In a moment I'll be back to summarize the program and offer some resources that can help you on your journey to secure your enterprise infrastructure. I wanna thank our guests for their contributions in helping us understand how investments by a company like Dell can both reduce the need for dev sec up teams to worry about some of the more fundamental security issues around infrastructure and have greater confidence in the quality providence and data protection designed in to core infrastructure like servers, storage, networking, and hyper-converged systems. You know, at the end of the day, whether your workloads are in the cloud, on prem or at the edge, you are responsible for your own security. But vendor r and d and vendor process must play an important role in easing the burden faced by security devs and operation teams. And on behalf of the cube production content and social teams as well as Dell Technologies, we want to thank you for watching a blueprint for trusted infrastructure. Remember part one of this series as well as all the videos associated with this program and of course today's program are available on demand@thecube.net with additional coverage@siliconangle.com. And you can go to dell.com/security solutions dell.com/security solutions to learn more about Dell's approach to securing infrastructure. And there's tons of additional resources that can help you on your journey. This is Dave Valante for the Cube, your leader in enterprise and emerging tech coverage. We'll see you next time.

>> From the cube studios in Palo Alto in Boston bringing you data-driven insights from the cube in ETR. This is breaking analysis with Dave Vellante. >> There are four A players, in the IS slash pass hyperscale cloud services space, AWS, Azure, Alibaba, and alphabet, pretty clever, huh? In our view, these four have the resources, the momentum, and stamina to outperform all others virtually indefinitely. Now combined, we believe these companies will generate more than $115 billion in 2021 IaaS and PaaS revenue. That is a substantial chunk of market opportunity that is growing as a whole in the mid 30% range in 2021. Welcome to this week's Wiki bond cube insights, powered by ETR. In this breaking analysis, we are initiating coverage of Alibaba for our IaaS and PaaS market segments. And we'll update you on the latest hyperscale cloud market data, and survey data from ETR. Big week in hyperscale cloud land, Amazon and alphabet reported earnings and AWS CEO Andy Jassy was promoted to lead Amazon overall. I interviewed John Furrier on the cube this week. John has a close relationship with Jassy and a unique perspective on these developments. And we simulcast the interview on clubhouse, and then hosted a two hour clubhouse room that brought together all kinds of great perspectives on the topic. And then, we took the conversation to Twitter. Now in that discussion, we were just riffing on our updated cloud estimates and our numbers. And here's this tweet that inspired the addition of Alibaba. Now this gentleman is a tech journalist out of New Delhi and he pointed out that we were kind of overlooking Alibaba and I responded that no, we do not just discounting them but we just need to do more homework in the company's cloud business. He also said we're ignoring IBM, but really they're not in this conversation as a hyperscale IaaS competitor to the big four in our view. And we'll just leave it at that for now on IBM, but, back to Alibaba and the big four, we actually did some homework. So thank you for that suggestion. And this chart shows our updated IaaS figures and includes the full year 2020 which was pretty close to our Q4 projections. You know, the big change is we've added Alibaba in the mix. Now these four companies last year, accounted for $86 billion in revenue, and they grew it 41% rate combined relative to 2019. Now, notably as your revenue for the first time is more than half of that of AWS's revenue which of course hit over $45 billion. AWS's revenue, over top 45 billion last year, which is just astounding. Alibaba you'll note, is larger than Google cloud. The Google cloud platform, I should say GCP, at just over eight billion for Alibaba. Now, the reason Baba is such a formidable competitor, is because the vast majority of its revenue comes from China inside that country. And the company do have plans to continue their international expansion, so we see Alibaba as a real force here. Their cloud business showed positive EBITDA for the first time in the history of the company last quarter. So that has people excited. Now, Google, as we've often reported, is far behind AWS and Azure, despite its higher growth rates Google's overall cloud business lost 5.6 billion in 2020 which has some people concerned. We on the other hand are thrilled, because as we've reported in our view, Google needs to get its head out of its ads cloud is it's future. And we're very excited about the company pouring investments into its cloud business. Look with $120 billion essentially in the balance sheet, we can think of a better use of its cash. Now, I want to stress that these figures are our best efforts to create an apples to apples comparison across all four clouds. Many people have asked about, how much of these figures represent, for example, Microsoft office 365 or Google G suite, which by the way now is called workspaces. And the answer is our intention is $0. These are our estimates of worldwide IaaS in PaaS revenue. You know, some of said, we're too low. Some of said, we're too high. Hey, if you have better numbers, Please share them, happy to have a look. Now you maybe asking, what are the drivers of these figures and the growth that we're showing here? Well, all four of these companies, of course, they're benefiting from an accelerated shift to digital as a result to COVID, but each one has other tailwinds. You know, for example, AWS, it's Capitalizing on its a large headstart. It's created tremendous brand value. And as well, despite the fact that, while we estimate that more than 75% of AWS revenue comes from compute and storage, AWS is feature and functional differentiation combined with this large ecosystem is a very much a driving force of it's growth. In the case of Azure, in addition to its captive software application estate, the company on its earnings calls cited strong growth in its consumption based business across all of its industries and customer segments. As we've said, many times, Microsoft makes it really easy for customers to tap into Azure and a true consumption pricing model, with no minimums and cancel any time. Those kinds of terms make it extremely attractive to experiment and get hooked. We certainly saw this with AWS over the years. Now for Google it's growth is being powered by its outstanding technology, and in particular its prowess in AI and analytics. As well we suspect that much of the losses in Google cloud are coming from large go to market investments for Google cloud platform, and they're paying growth dividends. Now, as Tim Crawford said on Twitter, 6 billion, you know that's not too shabby. Also Google cited wins at Wayfair in Etsy, that Google is putting forth in our view to signal that many retailers they might be are you reluctant to do business with Amazon, was of course a big retailer competitor. These are two high profile names, we'd like to see more in future quarters and likely will. Now let's give you another view of this data and paint a picture of, how the pie is being carved out in the market. Actually we'll use bars because my, millennials sounding boards they hate pie charts. And I like to pay attention, to these emerging voices. At any rate amongst these four, AWS has more than half of the market. AWS and Azure are well ahead of the rest. And we think we'll continue to hold serve for quite some time. Now while we're impressed with Alibaba, they're currently constrained to doing business mostly in China. And we think it'll take many years for Baba and GCP to close that gap on the two leaders if they'll ever even get there. Now let's take a look at, what the customers are saying within the ETR survey data. The chart that we're showing here, this is X, Y chart that we show all the time. It's got net score or spending moments on the vertical axis, and market share or the pervasiveness in the datasets in the survey on the horizontal axis. Now on the upper right, you can see the net scores and the number of mentions for each company and the detailed behind this data. And what we've done here is cut the January survey data of 1,262 respondents, you can see that in filtered in there on the left, and we've filtered the data by cloud meaning the respondents are answering about the companies, cloud computing offerings only. So we're filtering out anything of the non-cloud spend. That's a nice little capability of the ETR platform. Azure is really quite amazing to us. It's got a net score of 72.6%, and that's across 572 responses out of the 1262. AWS is the next most pervasive in the data set with 492 shared accounts and a net score of 57.1%. Now, you may be wondering, well, why is Azure bigger in the dataset than AWS? And when we just told you that the opposite is the case in the market in the previous slide. And the answer is, like this is a survey and it's a lot of Microsoft out there, they're everywhere. And I have no doubt that the respondants notion of cloud doesn't directly map into IaaS and PaaS views of the world, but the trends are clear and consistent. Amazon and Azure, they dominate in this market space. Now for context, we've included functions in the form of AWS Lambda as your functions and Google cloud functions. Because, as you can see, there's a lot of spending momentum in these capabilities in these services. You'll also note, that we've added Alibaba to this chart, and it's got a respectable 63.6% net Score, but there are only 11 shared responses in the data. So they'll go into the bank on these numbers, but look, 11 data points, we'll take it. It's better than zero data points. We've also added VMware cloud on AWS in this chart, and you can see that, that capability that service, that has the momentum and you can see those ones that we've highlighted above the 40% red dotted line, that's where the real action in the market is. So all of those offerings have very strong or strong spending velocity in the ETR data set. Now, for context, we've put Oracle and IBM in the chart. And you can see, they both have, you know they've got a decent presence in the data set. They have 132 mentions and 81 responses respectively. So Oracle, they've got a positive net score of 16.7%, and IBM is in a negative 6.2%. Now, remember this is for their cloud offerings, as the respondents in the data set see them. So what does this mean? It says that among the 132 survey respondents answering that they use Oracle cloud, 16.7% more customers are spending more on Oracle's cloud than are spending less. In the case of IBM, it says more customers are spending less than spending more. Both companies are in the red zone, and show far less momentum than the leaders. Look, I've said many times that the good news is, that Oracle and IBM at least have clouds. But they're not direct competitors of the big four in our view, there just not. They have a large software business, and they can migrate their customers, to their respective clouds and market hybrid cloud services. Their definition of cloud is most certainly different than that of AWS, which is fine, but both companies use what I call a kitchen sink method of reporting their cloud business. Oracle includes, cloud and license support, often with revenue recognition at the time of contract, With a term that's renewable and, it also includes on-prem fees, for things like database and middleware, and if, you want to call that cloud, fine. IBM is just as bad, maybe they're worse and includes so much legacy stuff and its cloud number to hide the ball. It's just not even worth trying to unpack for this episode, I have previously and frankly, it's just not a good use of time. Now, as I've said before, both companies they're in the game that can make good money provisioning infrastructure to support their respective software businesses. I just don't consider them hyperscale class clouds which are defined by the big four, and really only those four. And I'm sure I'll get hate mail about that statement, and I'm happy to defend that position, so please reach out. Okay, but one other important thing that we want to discuss is something that came up this week in our Twitter conversation. Here's a tweet from Matt Baker who had strategic planning for Dell. He was responding to someone who commented on our cloud data, basically saying that, with all that cloud revenue who took the hit, which pockets did it come out of, and Matt was saying, look, it's coming out of customer pockets, but can we please end this zero sum game narrative. In other words, it's not a dollar for cloud that doesn't translate into a lost dollar from on-prem for the legacy companies. So let's take a look at that. For first I would agree, with Matt Baker, it's not a one for one swap of spend but there's definitely been an impact. And here's some data from ETR that can, maybe give us some insight here. What this chart shows is a cut of 915 hyperscale cloud accounts. So within those big four, and within those accounts we show the spending velocity or net score cut within further sectors representative of these on-prem players. So servers, storage and networking, so we cut the data on those three segments. And we're looking here at, VMware, Cisco, Dell, HPE, and IBM, for 2020 and into 2021. It's kind of an interesting picture, it shows the net scores for the January of 20 April, July and October 20 surveys and the January 21 surveys. Now all the on-prem players, they were of course impacted by COVID, IBM seems to be that counter trend line. Not that they weren't impacted, but they have this notable mainframe cycle thing going on. And you know, they're in a down cycle now. So it's kind of opposite of the other guys in terms of the survey momentum. And you can see pretty much, all the others are showing upticks headed into 2021, Cisco, you know kind of flattish, but stable and held up a bit. So to Matt Baker's point, despite the 35% or so growth expected for the big four and 2021 the on-prem leaders are showing some signs of positive spending momentum. So let's dig into this a little bit further, 'cause we're not saying cloud hasn't hurt on prem spending. You know, of course it has. Here's that same picture, over a 10 year view. So you're seeing this long, slow, decline occur, and it's no surprise. If you think about the prevailing model for servers, storage, and networking, on prem in particular. Servers have been perpetually under utilized, even with virtualization. You know, with the exception of like backup jobs, there aren't many workloads that can max out server utilization. So we kept buying more servers to give us performance headroom and ran at 20, 30% utilization, you know in a good day. Yes I know some folks can get up over 50%, but generally speaking servers are well under utilized in storage my gosh, it's kind of the same story, maybe even worse. Because for years it was powered by a mechanical system. So more spindles are required to gain performance, lots of copying going on, lots of, you know, pre-flash waste. And in networking it was a story of got to buy more ports. You've got to buy more ports. In the case of these segments, customers will just defense essentially, forced in this endless cycle of planning, procuring, you know, first planning. They got to get the secure the CapEx, and then they procure, and then they over-provision, and then they manage, you know, ongoing. So then along comes AWS, and says, try this on for size and you can see from that chart, the impact of cloud on those bellwether on-prem infrastructure players. Now, just to give you a little bit more insight on this topic, here's a picture of the wheel charts from the ETR data set. For AWS Microsoft, Google, and we brought in VMware to compare them. A wheel chart shows the percent of customers saying they'll either add a platform new that's the lime green. Increased spending by more than 5%, that's the forest green spend flat relative to last year. That's the gray spend less by more than 5% down, that's the pinkish or leave the platform, that's the Bright red. You subtract the red from the green and you get a percentage that represents net score, AWS with a net score of 60% is off the charts good. Microsoft remember, this includes the entire Microsoft business portfolio, not just Azure, so it's still really strong. Google, frankly, we'd like to see higher net scores and VMware's, you know, so there's a gold standard for on-prem. So we include them, so you can see for reference the strong, but notice they got a much, much bigger flat spending, which is what you would expect from some of these more mature players. Now let's compare these scores to the other, on-prem Kings. So this is not surprising to see, but the greens, they go down, the flats that gray area goes up compared to the cloud guys and the red which is virtually non-existent within AWS, goes into the high teens with the exception of Cisco which despite its exposure to virtually all industries including those hard hit by COVID shows pretty low read scores. So that's, that's good. And I got to share one other, look at this wheel chart for pure storage. We're not really not sure what's happening here, but this is impressive. We're seeing a huge rebound, and you can see we've superimposed as candlestick over comparing previous quarters surveys and, look at the huge up check in the January survey for pure that blue line. That's highlighted in that red dot at ellipse, jumps to a 63% net score from below 20% last quarter. You know, we'll see, I've never seen that kind of uptick before for an established company. And, you know, maybe it's pent up demand or some other anomaly in the data. We'll find out when pure reports in 2021, because remember these are forward looking surveys. But the point is, you still see action going on in hybrid and on-prem, and despite the freight train that is cloud, coming at the legacy players. You know, not that pure is legacy, but it's, you know, it's no longer a lanky teenager. And I think the bottom line, coming back to Matt Baker's point, is there are opportunities that the on-prem players can pursue in hybrid and multi-cloud, and we've talked about this a lot where you're building abstraction layer, on top of the hyperscale clouds and letting them build out their data center presence worldwide, spend on capex, they're going to outspend everybody. And these guys, these on-prem, and hybrid and multi-cloud folks they're going to have to add value on top of that. Now if they move fast, you no doubt there'll be acquiring startups to make that happen. They're going to have to put forth the value proposition and execute on that, in a way that adds clear value above and beyond what the hyperscalers are going to do. Now, the challenge, is picking those right spots, moving fast enough and balancing wall street promises with innovation. There's that same old dilemma. Let's face It. Amazon for years could lose tons of money and not get killed in the street. Google, they got so much cash, they can't spend it fast enough and Microsoft after years of going sideways is finally figured out and the some. Alibaba they're new to our analysis, but it's looking like you know, it's the Amazon of China, Plus ANT despite its regulatory challenges with the Chinese government. So all four of these players, are in the driver's seat in our view. And they're leading in not only cloud, but AI. And of course the data keeps flowing into their cloud. So they're really are in a strong position. Bottom line is we're still early into the cloud platform era and it's morphing. It's from a collection of remote cloud services, into this ubiquitous, sensing, thinking, anticipatory system, that's increasingly automated and working towards full automation. It's intelligent and it's hyper decentralizing toward the edge. One thing's for sure, the next 10 years, they're not going to be the same as the past 10. Okay, that's it for now. Remember I publish each week on Wikibond.com and siliconANGLE.com, these episodes they're all available as podcasts just search for breaking analysis podcast. You can always connect on Twitter. I'm @dvellante or email me at david.Vellante@siliconANGLE.com. I love the comments on LinkedIn and of course in clubhouse the new social app. So please follow me, so that you can get notified when we start a room and riff on these topics. And don't forget to check out etr.plus for all the survey action. This is Dave Vellante for the cube insights powered by ETR be well, and we'll see you next time. (upbeat music)

>> From theCube studios in Palo Alto in Boston, bringing you data-driven insights from theCube in ETR. This is Breaking Analysis with Dave Vellante. >> New data suggests the tech spending will be higher than we previously thought for 2021. COVID learnings, a faster than expected vaccine rollout, productivity gains in the last 10 months, and broad-based cloud leverage lead us to raise our outlook for next year. We now expect a three to 5% increase in 2021 technology spending, roughly double our previously forecasted growth rate of 2%. Hello everyone and welcome to this week's we keep on Cube Insights powered by ETR. In this breaking analysis, we're going to share new spending data from ETR partners and take a preliminary look at which sectors and which companies are showing momentum heading into next year. Let's get right into it. The data is pointing to a strong 2021 rebound. A latest survey from ETR and the information from theCube Community suggests that the accelerated pace of the vaccine rollout pent up demand for normalcy and learnings from COVID will boost 2021 tech spending higher than previously anticipated. Now a key factor we've cited is that the forced March to digital transformation due to the pandemic created a massive proof of concept for what works and what doesn't in a digital business. CIOs are planning to bet on those sure things to drive continued productivity improvements and new business opportunities. Now, speaking of productivity, nearly 80% of respondents in the latest ETR survey indicate that productivity either stayed the same or improved over the past three months. Now of those, the vast majority, more than 80% cited improvements in productivity. This has been a common theme throughout the year. As well, the expectation among CIOs is that many workers will return to the office in the second half of the year, which we expect will drive new spending in the infrastructure needs of company HQs, which have been neglected over the past 10 months. Now, despite the expectation that many workers will return to the office, 2020 has shown us that working remotely, hey, it's here to stay, and a much larger number of employees are going to be permanently remote working than pre pandemic. ETR survey data shows that that number is going to be approximately double over the longterm. We'll look at some of that specific data. In addition, cloud computing, it became the staple of business viability in 2020. Those that were up the cloud adoption ramp, well, they benefited greatly, those that weren't well, they had to learn fast. Now, along with remote work cloud necessitated new thinking around network security, and as we've reported identity access management, endpoint security and cloud security with the beneficiaries. Companies like Okta, CrowdStrike, Zscaler, a number of others continue to ride this wave. Larger established security companies like Cisco, Palo Alto Networks, F5, Fortunate and others, they have major portions of their business that are benefiting from the tailwinds in the shift and network traffic, as a result of cloud and remote work. Now, despite all the momentum in the market and the expect of improvements in 2021, these tailwinds are not expected to be evenly distributed, far from it. We think Q4 is going to remain soft relative to last year and Q1 2021 is going to be flat, maybe up slightly. Remember the COVID impact was definitely felt in March of this year. So based on the earnings that we saw, there may be some upside in Q1, given that organizations are still being cautious in Q4, and really there's still some uncertainty in Q1. Let's look at some of the survey responses and you'll see why we're more optimistic than we've previously reported. This chart shows the responses to key questions around spending trajectories from the March, June, September, and December surveys of this year. Now it's no surprise that there's been little change in remote workers and limiting business travel. But look at the other categories, seeing a dramatic reduction in hiring freezes. The percentage of companies freezing new IT deployments continues to drop throughout the year. And then conversely, the percentage of companies accelerating new it deployments that's sharply up to 34% from the March low of 12%. And look at the headcount trends. The percentage of companies instituting layoffs. It continues its downward trajectory while accelerated hiring is now up to 17%. So there's a lot to be excited about in these results. Now let's look the remote worker trend. How do CIO see that shift in the near to midterm? This chart shows the work from home data and it's amazingly consistent from the September survey drill down. You can see CIO's is indicate that on average, 15 to 60% of workers were remote prior to the pandemic, and that jumped up to 72 to 73% currently, and is expected to stay in the high fifties until the summer of 2021. Thereafter, organizations expect that the number of employees that work remotely on a permanent basis is going to more than double to 34% long term. By the way, I've talked to a number of executives, CEOs, CIOs, and CFOs that expect that number to be higher than these especially in the technology sector. They expect more than half of their workers to be remote and are looking to consolidate facilities cost to save money. As we've said, cloud computing has been the most significant contributor to business resilience and digital transformation this year. So let's look at cloud strategies and see how CIOs expect those to evolve. This chart shows responses to how organizations see multi-cloud evolving. It's interesting to note the ETR call-out, which concludes that the narrative around multi-cloud multi-cloud is real, and it is. But I want to talk to you about a flip side to this notion in that, as many customers have, or are planning to increasingly concentrate workloads in the cloud. This actually makes some sense. Sure, virtually every major company uses multiple clouds, but more often than not, it concentrate work on a primary cloud. CIO strategies, they're not generally evenly distributed across clouds. The data shows that this is the case for less than 20% of the respondents, rather organizations are typically going to apply an 80, 20 or a 70, 30 rule for their multi-cloud approach. Meaning they pick a primary cloud on which most work is done, and then they use alternative clouds as either a hedge or maybe for specific workloads or maybe even data protection purposes. Now, if you think about it, optimizing on a primary cloud allows organizations to simplify their security and governance and consolidate their skills. At this point in the cloud evolution, it seems CIOs feel there's more value that is going to come from leveraging the cloud to change their operating models, and maybe broadly spreading the wealth to reduce risk or maybe cut costs, or maybe even to tap specialized capabilities. What's more in thinking about AWS and Microsoft respectively. Each can make a very strong case from MANO cloud. AWS has more features than any other cloud, and as such can handle most workloads. Microsoft can make a similar argument for its customers that have an affinity and a largest state of Microsoft software. The key for multi-cloud in our view will be the degree to which technology vendors can abstract the underlying cloud complexity and create a layer that floats above the clouds and adds incremental value. Snowflakes data cloud is one of the best examples of this, and we've covered that pretty extensively. Now, clearly VMware and Red Hat have aspirations at the infrastructure layer in a similar fashion. Pure storage, and NetApp are a couple of the largest storage players with similar visions. And then Qumulo and Clumio are two other examples with promising technologies, but they have a much smaller install base. Take a look at Cisco, Dell, IBM and HPE. They have a lot to gain and a lot to lose in this cloud game. So multi-cloud is an imperative for these leaders, but for them it's much more complicated because of the complexity and vastness of their portfolios. And notably Dell has VMware and IBM of course has Red Hat, which are key assets that can be leveraged for this multi-cloud game. HPE has a channel and a large install base, but all of these firms, they have to spread R&D much more thinly than some of these other companies that we mentioned for example. The bottom line is that multi-cloud has to be more than just plugging into an operating well on any of the clouds. It require... Which is by the way, this is mostly where we are today. It requires an incremental value proposition that solves a clear problem, and at the same time runs efficiently, meaning it takes advantage of cloud native services at scale. What sectors are showing momentum heading into 2021? And who are some of the names that are looking strong? We've reported a lot that cloud containers and container orchestration, machine intelligence and automation are by far the hottest sectors, the biggest areas of investment with the greatest spending momentum. Now we measure this in ETR parlance, remember by net score. But here's the good news, almost every other sector in the ETR taxonomy with the notable exception of IT outsourcing and IT consulting is showing positive spending momentum relative to previous surveys this year. Yeah, maybe not, it's not a shock, but it appears that the tech spending recovery will be broad-based. It's also worth noting that there are several vendors that stand out and we show a number of them here. CrowdStrike, Microsoft has had consistent performance in the dataset throughout this year. Okta, we called out those guys last year and they've clearly performed as you can see in their earnings reports. Pure storage, interestingly, big acceleration and a turnaround from last quarter in the dataset, and of course, snowflake has been off the charts as we reported many times. These guys are all seeing highly accelerated momentum. UiPath just announced its intent to IPO, AWS, Google, Zscaler, SailPoint, ServiceNow, and Elastic, these all continue to trend up. And so, there are some real positives that we're looking for a member of the ETR surveys, they're forward-looking. So we'll see, as we catch up next quarter. Now, before we wrap, I want to say a few words on security, and maybe it's a bit of a non-sequitur here, but I think it's relevant to the trends that we've been discussing, especially as we talk about moving to the cloud. And as you know, we've reported many times on the security space, basically updating you quarterly with our scenarios and the spending and the technology trends and highlighting our four-star companies. Four-star company's insecurity on those with both momentum and significant market presence. And last year we put CrowdStrike, Okta and Zscaler, and some others on the radar. And we've closely track the cyber business of larger companies with a security portfolio like Palo Alto and Cisco, and more recently, VMware has made some acquisitions. Now the government hacked that became news this week. It really underscores the importance of security. It remains the most challenging area for organizations because well, failure's not an option, skills are short, tools are abundant, the adversaries are very well-funded and extremely capable yet failure is common as we saw this week. And there's a misconception that cloud solves the security problem, and it's important to point out that it does not. Cloud is a shared responsibility model, meaning the cloud provider is going to secure the infrastructure for example, but it's up to you as the customer to configure things properly and deal with application security. It's ultimately on you. And the example of S3 is instructive because we've seen a number S3 breaches over the years where the customer didn't properly configure the S3 bucket. We're talking about companies like Honda and Capital One, not just small businesses that don't have the SecOps resources. And generally it was because a non-security person was configuring things. Maybe they were Or developers who are not focused on security, and perhaps permission set too broadly, and access was given to far too many people. Whatever the issue, it took some breaches and subsequent education to increase awareness of this problem and tighten it up. We see some similar trends occurring with new workloads, especially in cloud databases. It's becoming so easy to spin up new data warehouses for example, and we believe that there are exposures out there due the lack of awareness or inconsistent corporate governance being applied to these new data stores. As well, even though important areas like threat intelligence and database security are important, SecOps budgets are stretched thin. And when you ask companies where the priorities are, these fall lower down the list, these areas specifically have taken a back seat, the endpoint, identity and cloud security. And we bring this up because it's a potential blind spot as we saw this week with the US government hack. It was stealthy, it wasn't detected for many, many months. Who knows maybe even years. And not to be a buzzkill, but the point is, cloud enthusiasm has to be concompetent with security vigilant. Enough preaching, let's wrap up here. As we enter 2020, this year, we said the cloud was going to be the force that drove innovation along with data and AI. And as we look in the rear view mirror and put 2020 behind us, I know many of you want to do that, it was the cloud that enabled businesses to not only continue to operate, but to actually increase productivity. Nonetheless, we still see IT spending declines of four to 5% this year with an expectation of a tepid Q4 relative to the last year. We see Q1 slowly rebounding and kind of a swoosh, let me try that again, recovery in the subsequent quarters with tech spending rebounding in 2021 to a positive three to 5%, let's call it 4%. Now supporting us scenario, the pandemic forced a giant Petri dish for digital. And we see some real successes and learnings that organizations will apply in 2021 to bet on sure things. These are cloud, containers, AI, ML, machine intelligence pieces and automation. For sure, along with upticks for virtually every other sector of technology because spending has been so depressed. The two exceptions are outsourcing and IT consulting and related services which continue to be a drag on overall spending. Priorities must be focused on security and governance and further improvements in applying corporate edicts in a cloud world. We also see new data architectures emerging where domain knowledge becomes central to data platforms. We'll be covering this in more detail on top of the work that we've already done in this area. Now, automation is not only an opportunity, it's become a mandate. Yes, RPA, but also broader automation agendas be on point tools. And importantly, we're not talking about paving the cow path here by automating existing processes. Rather we're talking about rethinking processes across the entire organization for a new digital reality where many of these processes are being invented. The work of Erik Brynjolfsson and Andrew McAfee on the second machine age. It was pressured back in 2014 and the conclusions they drew, they're becoming increasingly important in the 2020s, meaning that look machines have always replaced humans throughout time. But for the first time in history, it's happening for cognitive functions, and a huge base of workers is going to be, or as being marginalized, unless they're retrained. Education and public policy that supports this transition is critical. And I for one would like to see a much more productive discussion that goes beyond the cult of break up big tech. Rather I'd like to see governments partner with big tech to truly do good and help drive the re-skilling of workers for the digital age. Now cloud remains the underpinning of the digital business mandate, but the path forward isn't really always crystal clear. This is evidenced by the virtual dead heat between those organizations that are consolidating workloads in a cloud workloads versus those that are hedging bets on a multi-cloud strategy. One thing is clear cloud is the linchpin for our growth scenarios and will continue to be the substrate for innovation in the coming decade. Remember, these episodes, they're all available as podcasts, wherever you listen, all you got to do is search Breaking Analysis podcast, and please subscribe to the series, appreciate that. Check out ETR's website at ETR.plus. We also publish full report every week on wikibond.com and siliconangle.com and get in touch with me at David.vallante, siliconangle.Com, you can DM me at D. Vellante. And please by all means comment on our LinkedIn posts. This is Dave Vellante for theCube Insights powered by ETR. Have a great week everybody, Merry Christmas, happy Hanukkah, happy Kwanzaa, or happy, whatever holiday you celebrate. Stay safe, be well, and we'll see you next time. (upbeat music)

>>Yeah. >>Welcome back for our last session of the day how to deliver career making business outcomes with Search and AI. So we're very lucky to be hearing from Canada. Canadian Tire, one of Canada's largest and most successful retailers, have been powered 4.5 1000 employees to maximize the value of data with self service insights. So today we're joining us. We have Yarrow Baturin, who is the manager of Merch analytics and planning to support at Canadian Tire and then also Andrea Frisk, who is the engagement manager manager for thoughts. What s O U R Andrea? Thanks so much for being here. And with >>that, >>I'll pass the mic to you guys. >>Thank you for having us. Um, already, I I think I'll start with an introduction off who I am, what I do. A Canadian entire on what Canadian pair is all about. So, as a manager of Merch analytics at Canadian Tire, I support merchant organization with reporting tools, and then be I platform to enable decision making on a day to day basis. What is? Canadian Tire's Canadian tire is one of the largest retailers in Canada. Um, serving Canadians with a number of lines of business spanning automotive fixing, living, playing and SNG departments. We have a number of banners, including sport check Marks Party City Phl that covers more than 1700 locations. So as an organization, we've got vast variety of different data, whether it's product or loyalty. Now, as the time goes on, the number of asks the number off data points. The complexity of the analysis has been increasing on banned traditional tools. Analytical tools such as Excel Microsoft Access do find job but start hitting their limitations. So we started on the journey of exploring what other B I platforms would be suitable for our needs. And the criteria that we thought about as we started on that journey is to make sure that we enable customization as well as the McCarthy ization of data. What does that mean? That means we wanted to ensure that each one of the end users have ability to create their own versions off the report while having consistency from the data standpoint, we also wanted Thio ensure that they're able to create there at hawks search queries and draw insights based on the desired business needs. As each one of our lines of business as each one of our departments is quite unique in their nature. And this is where thoughts about comes into play. Um, you checked off all the boxes? Um, as current customers, as potential customers, you will discover that this is the tool that allows that at hawks search ability within a matter of seconds and ability to visualize the information and create those curated pin boards for each one of the business units, depending on what the needs are. And now where? I guess well, Andrea will talk a little bit more about how we gained adoption, but the usage was like and how we, uh, implemented the tool successfully in the organization. >>Okay, so I actually used to work for Canadian tire on DSO. During that time, I helped Thio build training and engaging users to sort of really kick start our use cases. Andi, the ongoing process of adopting thought spot through Canadian Tire s 01 of the sort of reasons that we moved into using thought spot was there was a need Thio evolve, um, in order to see the wealth of data that we had coming in. So the existing reporting again. And this is this sort of standard thoughts bought fix is, um, it brings the data toe. Everyone on git makes it more accessible, so you get more out of your data. So we want to provide users with the ability to customize what they could see and personalized three information so that they could get their specific business requirements out of the data rather than relying on the weekly monthly quarterly reporting. That was all usually fairly generic eso without the ability to deep dive in. So this gave the users the agility thio optimize their campaigns, optimize product murder, urgency where products are or where there's maybe supply chain gaps. Andi just really bring this out for trillions of rose to become accessible. Thio the Canadian tire. That's what user base think. That's the slide. >>That's the slight, Um So as Andrea talked about the business use of the particular tool, let's talk a little bit about how we set it up and a wonderful journey of how it's evolved. So we first implemented 5.3 version of that spot on the Falcon server on we've been adding horsepower to it over time. Now mhm. What I want to stress is the importance off the very first, Data said. That goes into the tool toe. Actually engage the users and to gain the adoption and to make sure there is no argument whether the tool is accurate or not. So what we've started with is a key p I marked layer with all the major metrics that we have and all the available permutations and combinations off the dimensions, whether it's a calendar dimension, proud of dimension or, let's say, customer attribute now, as we started with that data set, we wanted to make sure that we're we have the ability to add and the dimensions right. So now, as we're implementing the tool, we're starting to add in more dimension tables to satisfy the needs off our clients if you want to call it that way as they want to evolve their analytics. So we started adding in some of the store attributes we started adding in some of the product attributes on when I refer to a product attributes, let's say, uh, it involves costs and involves prices involved in some of the strategic internal pieces that we're thinking about now as the comprehensive mark contains right now, in our instance, close to five billion records. This is where it becomes the one source of truth for people declaring information against right so as they go in, we also wanted to make sure when they Corey thought spot there, we're really Onley. According one source of data. One source of truth. It became apparent over time, obviously, that more metrics are needed. They might not be all set up in that particular mark. And that's when we went on the journey off implementing some of the new worksheets or some of the new data sets particularly focused on the four looking pieces. And uh, that's where it becomes important to say This is how you gain the interest and keep the interests of the public right. So you're not just implementing a number off data sets all at once and then letting the users be you're implementing pieces and stages. You're keeping the interest thio, the tool relevant. You're keeping, um, the needs of the public in mind. Now, as you can imagine on the Falcon server piece, um, adding in the horsepower capacity might become challenging the mawr. Billions of Rosie erratic eso were actually in the middle of transitioning our environment to azure in snowflake so that we can connect it. Thio embrace capability of thoughts cloud. And that's where I'm looking forward to that in 2021 I truly believe this will enable us Thio increase the speed off adoption Increase the speed of getting insights out of the tool and scale with regards Thio new data sets that we're thinking about implementing as we're continuing our thoughts about journey >>Okay, so how we drove adoption Thio 4500 plus users eso When we first started Thio approach our use case with the merchants within Canadian Tire We had meetings with these users with who are used place is gonna be with and sort of found out. What are they searching for, Where they typically looking at what existing reports are available for them. Andi kind of sought out to like, What are those things where you're pulling this on your own or someone else's pulling this data because it's not accessible yet And we really use that as our foundation to determine one what data we needed to initially bring into the system but also to sort of create those launchpad pin boards that had the base information that the users we're gonna need so that we could twofold, make it easy for them, toe adopt into the tool and also quickly start Thio, deactivate or discontinue those reports. And just like these air now only available in thought spot because with the sort of formatting within thought spot around dates, it's really easy to make this year's report last year report etcetera. Just have everything roll over every month or a recorder s. So that was kind of some of the pre work foundation when we originally did it. But really, it's been a lot of training, a lot of training. So we conducted ah, lot of in person training, obviously pre co vid eso. We've started to train the group that we targeted, which was the merchants and all of the like, surrounding support groups. Eso we had planners going in and training as well, so that everyone who was really closely connected to the merchants I had an idea of what thoughts about what was and how to use it and where the reports were, and so we just sort of rolled it out that way, and then it started to fly like wildfire. Eso the merchants start to engage with supply chain to have conversations, or the merchants were engaging with the vendors to sort of have negotiations about pricing. And they're creating these reports and getting the access to the information so quickly, and they're sharing it out that we had other groups just coming to us asking, How do I get into thoughts about how can I get in on DSO on top of those groups, we also sought out other heavy analytics groups such a supply chain where we felt like they could have the same benefits if they on boarded into thought spot with their data as well on Ben. Just continuing to evolve the training roll out. Um, you know, we continued to engage with the users, >>so >>we had a newsletter briefly Thio, sort of just keep informing users of the new data coming in or when we actually upgraded our system. So the here are the new features that you'll start seeing. We did virtual trainings and maintaining an F A Q document with the incoming questions from the users, and then eventually evolved into a self guided learning so that users that were coming to a group, or maybe we've already done a full rollout could come in and have the opportunity to learn how to use thought spot, have examples that were relevant to the business and really get started. Eso then each use case sort of after our initial started to build into a formula of the things that we needed to have. So you need to understand it. Having SMEs ready and having the database Onda worksheets built out sort of became the step by step path to drive adoption. Um, from an implementation timeline, I think they're saying, Took about two months and about half of that waas Kenny entire figuring out how figuring out our security, how to get the data in on, Do we need the time to set up the environment and get on Falcon? So then, after that initial two months, then each use case that we come through. Generally, we've got users trained and SMEs set up within about 2 to 3 weeks after the data is ingested. It's not obviously, once snowflakes set up on the data starts to get into that and the data feeds in, then you're really just looking at the 2 to 3 weeks because the data is easily connected in, >>um, no. All right, let's talk about some of the use cases. So we started with what data we've implemented. Andrea touched upon what Use a training look like what the back curate that piece wants. Now let's talk a little bit about use cases and how we actually leverage thoughts bought together the insights. So the very first one is ultimately the benefit of the tool to the entire organization. Israel Time insights. To reiterate what Andrea said, we first implemented the tool with our buyers. They're the nucleus of any retail organization as they work with everybody within the company and as the buyer's eyes, Their responsibility to ensure both the procurement and the sales channel, um, stays afloat at the end of the day, right? So they need information on a regular basis. They needed fast. They needed timely, and they needed in a fashion that they choose to digest it. It right? Not every business is the same. Not every individual is the same. They consume digest, analyze information differently. And that's what that's what allows you to dio whether it's the search, whether it's a customized onboard, please now supply chain unexpected things. As Andrea mentioned Irish work a lot of supply chain. What is the goal of supply chain to receive product and to be able to ship that product to the stores Now, as our organization has been growing and is doing extremely well, we've actually published Q three results recently. Um, the aspect off prioritization at D C level becomes very important, And what drives some of that prioritization is the analysis around what the upcoming sales would be for specific products for specific categories. And that's where again thoughts. But is one of the tools that we've utilized recently to set our prioritization logic from both inbound and outbound us. It's right because it gives you most recent results. It gives you most granular results, depending on the business problem that you're trying to tackle. Now let's chat a little bit about covert 19 response, because this one is an extremely interesting case as a pandemic hit back in March. Um, as you can imagine, the everyday life a Canadian entire became as business unusual is our executives referred to it under business unusual. This speed and the intensity of the insights and the analytics has grown exponentially. And the speed and the intensity of the insights is driven by the fact that we were trying Thio ensure that we have the right selection of products for our Canadian customers because that's ultimately bread and butter off all of the retailers is the customers, right? So thoughts bought allowed us to have early trends off both sales and inventory patterns, where, whether we were stalking out of some of the products in specific stories of provinces, whether we saw some of the upload off different lines of business, depending on the region, ality right as pandemic hit, for example, um, gym's closed restaurants closed. So as Canadian pack carries a wide variety of different lines of business, we actually offer a wide selection of exercise equipment and accessories, cycling products as well as the kitchen appliances and kitchen accessories pieces. Right? So all of those items started growing exponentially and in certain areas more than others. And this is where thoughts about comes into play. A typical analysis on what the region ality of the sales has been over the last couple of days, which is lifetime and pandemic terms, um, could have taken days weeks for analysts to ultimately cobbled together an Excel spreadsheet. Meanwhile, it can take a couple of seconds for 12 Korean tosspot set up a PIN board that can be shared through a wide variety of individuals rather than fording that one Excel spreadsheet that gets manipulated every single time. And then you don't get the right inside. So from again merch supply chain covert response aspect of things. That spot has been one of those blessings and one of those amazing tools to utilize and improve the speed off insights, improved the speed of analytics and improve the speed of decision making that's ultimately impacting, then consumer at the store level. So Andrea talked about 4500 users that we have that number of school. But what I owe the recently like to focus on, uh, Andrew and I laughing because I think the last time we've spoken at a larger forum with the fastball community, I think we had only 500 users. That was in the beginning >>of the year in in February, we were aiming to have like 1000 >>exactly. So mission accomplished. So we've got 4500 employees now. Everybody asked me, Yeah, that's a big number, but how many times do people actually log in on a weekly or daily basis? I'm or interested in that statistic? So lately, um, we've had more than 400 users on the weekly basis. What's what's been cool lately is, uh, the exponential growth off ad hoc ways. So throughout October, we've reached a 75,000 ad hoc ways in our system and about 13,000 PIN board views. So why is that's that's significant? We started off, I would say, in January of 2020 when Andrea refers to it, I think we started off with about 40 45,000 ad hoc worries a month. So again, that was cool. But at the end of the day, we were able to thio double that amount as more people migrate to act hawk searches from PIN board views, and that's that's a tremendous phenomena, because that's what that's about is all about. So I touched upon a little bit about exercise and cycling. So these are our quarterly results for Q two, um, that have showed tremendous growth that we did not plan for, that we were able to achieve with, ultimately the individuals who work throughout the organization, whether it's the merch organization or whether it's the supply chain side of the business. But coming together and utilizing a B I platform by tools such a hot spot, we can see triple digit growth results. Eso What's next for us users at Hawks searches? That's fantastic. I would still like to get to more than 1200 people on the weekly basis. The cool number to me is if all of our lifetime users were you were getting into the tool on a weekly basis. That would be cool. And what's proven to be true is ultimately the only way to achieve it is to keep surprising and delighting them and your surprising and delighting them with the functionality of the tool. With more of the relevant content and ultimately data adding in more data, um, is again possible through ET else, and it's possible through pulling that information manually. But it's expensive, expensive not from the sense of monetary value, but it's expensive from the size time, all of those aspects of things So what I'm looking forward to is migrating our platform to azure in snowflake and being able thio scale our insights accordingly. Toe adding more data to Adam or incites more, uh, more individual worksheets and data sets for people to Korea against helps the each one of the individuals learn. Get some of the insights. Helps my team in particular be, well, more well versed in the data that we have existing throughout the organization. Um, and then now Andrea, in touch upon how we scale it further and and how each one of the individuals can become better with this wonderful >>Yeah, soas used a zero mentioned theater hawk searches going up. It's sort of it's a little internal victory because our starting platform had really been thio build the pin boards to replicate what the users were already expecting. So that was sort of how we easily got people in. And then we just cut off the tap Thio, whatever the previous report waas. So it gave them away. Thio get into the tool and understand the information. So now that they're using ad hoc really means they understand the tool. Um, then they they have the data literacy Thio access the information and use it how they need. So that's it's a really cool piece. Um, that worked on for Canadian tire. A very report oriented and heavy organization. So it was a good starting platforms. So seeing those ad hoc searches go up is great. Um, one of the ways that we sort of scaled out of our initial group and I kind of mentioned this earlier I sort of stepped on my own toes here. Um is that once it was a proven success with the merchants and it started to spread through word of mouth and we sought out the analyst teams. Um, we really just kept sort of driving the insights, finding the data and learning more about the pieces of the business. As you would like to think he knows everything about everything. He only knows what he knows. Eso You have to continue to cultivate the internal champions. Um Thio really keep growing the adoption eso find this means that air excited about the possibility of using thought spot and what they can do with it. You need to find those people because they're the ones who are going to be excited to have this rapid access to the information and also to just be able to quickly spend less time telling a user had access it in thought spot. Then they would running the report because euro mentioned we basically hit a curiosity tax, right? You you didn't want to search for things or you didn't want to ask questions of the data because it was so conversed. Um, it was took too much time to get the data. And if you didn't know exactly what you were looking for, it was worse. So, you know, you wouldn't run a query and be like, Oh, that's interesting. Let me let me now run another query of all that information to get more data. Just not. It's not time effective or resource effective. Actually, at the point, eso scaling the adoption is really cultivating those people who are really into it as well. Um, from a personal development perspective, sort of as a user, I mean, one who doesn't like being smartest person in the room on bought spot sort of provides that possibility. Andi, it makes it easier for you to get recognized for delivering results on Dahlia ble insights and sort of driving the business forward. So you know, B b that all star be the Trailblazer with all the answers, and then you can just sort of find out what really like helping the organization realized the power of thought spot on, baby. Make it into a career. >>Amazing. I love love that you've joined us, Andrea. Such a such an amazing create trajectory. No bias that all of my s o heaps of great information there. Thank you both. So much for sharing your story on driving such amazing adoption and the impact that you've been able to make a T organization through. That we've got a couple of minutes remaining. So just enough time for questions. Eso Andrea. Our first questions for you from your experience. What is one thing you would recommend to new thoughts about users? >>Um, yeah, I would say Be curious and creative. Um, there's one phrase that we used a lot in training, which was just mess around in the tool. Um, it's sort of became a catchphrase. It is really true. Just just try and use it. You can't break. It s Oh, just just play around. Try it you're only limitation of what you're gonna find is your own creativity. Um, and the last thing I would say is don't get trapped by trying to replicate things. Is that exactly as they were? B, this is how we've always done it. Isin necessarily The the best move on day isn't necessarily gonna find new insights. Right. So the change forces you thio look at things from a different perspective on defined. Find new value in the data. >>Yeah, absolutely. Sage advice there. Andan another one here for Yaro. So I guess our theme for beyond this year is analytics meets Cloud Open for everyone. So, in your experience, what does What does that mean for you? >>Wonderful question. Yeah. Listen, Angela Okay, so to me, in short, uh, means scale and it means turning Yes. Sorry. No, into a yes. Uh, no, I'm gonna elaborate. Is interest is laughing at me a little bit. That's right. >>I can talk >>Fancy Two. Okay, So scale from the scale perspective Cloud a zai touched upon Throw our conversation on our presentation cloud enables your ability Thio store have more data, have access to more data without necessarily employing a number off PTL developers and going toe a number of security aspect of things in different data sources now turning a no into a yes. What does that mean with more data with more scalability? Um, the analytics possibilities become infinite throughout my career at Canadian Tire. Other organizations, if you don't necessarily have access thio data or you do not have the necessary granularity, you always tell individuals No, it's not possible. I'm not able to deliver that result. And quite often that becomes the norm, saying no becomes the norm. And I think what we're all striving towards here on this call Aziz part the conference is turning that no one say yes on then making a yes a new, uh, standard a new form. Um, as we have more access to the data, more access to the insights. So that would be my answer. >>Love it. Amazing. Well, that kind of brings in into this session. So thank you, everyone for joining us today on did wrap up this dream. Don't miss the upcoming product roadmap eso We'll be sticking around to speak thio some of the speakers you heard earlier today and I'll make the experts round table, and you can absolutely continue the conversation with this life. Q. On Q and A So you've got an opportunity here to ask questions that maybe keep you up at night. Perhaps, but yet stay tuned for the meat. The experts secrets to scaling analytics adoption after the product roadmap session. Thanks everyone. And thank you again for joining us. Guys. Appreciate it. >>Thank you. Thanks. Thanks.

>>from around the globe. It's the Cube presenting adaptive data governance brought to you by Iota Ho. >>And we're back with the data automation. Siri's. In this episode, we're gonna learn more about what I owe Tahoe is doing in the field of adaptive data governance how it can help achieve business outcomes and mitigate data security risks. I'm Lisa Martin, and I'm joined by a J. Bihar on the CEO of Iot Tahoe and Lester Waters, the CEO of Bio Tahoe. Gentlemen, it's great to have you on the program. >>Thank you. Lisa is good to be back. >>Great. Staley's >>likewise very socially distant. Of course as we are. Listen, we're gonna start with you. What's going on? And I am Tahoe. What's name? Well, >>I've been with Iot Tahoe for a little over the year, and one thing I've learned is every customer needs air just a bit different. So we've been working on our next major release of the I O. Tahoe product. But to really try to address these customer concerns because, you know, we wanna we wanna be flexible enough in order to come in and not just profile the date and not just understand data quality and lineage, but also to address the unique needs of each and every customer that we have. And so that required a platform rewrite of our product so that we could, uh, extend the product without building a new version of the product. We wanted to be able to have plausible modules. We also focused a lot on performance. That's very important with the bulk of data that we deal with that we're able to pass through that data in a single pass and do the analytics that are needed, whether it's, uh, lineage, data quality or just identifying the underlying data. And we're incorporating all that we've learned. We're tuning up our machine learning we're analyzing on MAWR dimensions than we've ever done before. We're able to do data quality without doing a Nen initial rejects for, for example, just out of the box. So I think it's all of these things were coming together to form our next version of our product. We're really excited by it, >>So it's exciting a J from the CEO's level. What's going on? >>Wow, I think just building on that. But let's still just mentioned there. It's were growing pretty quickly with our partners. And today, here with Oracle are excited. Thio explain how that shaping up lots of collaboration already with Oracle in government, in insurance, on in banking and we're excited because we get to have an impact. It's real satisfying to see how we're able. Thio. Help businesses transform, Redefine what's possible with their data on bond. Having I recall there is a partner, uh, to lean in with is definitely helping. >>Excellent. We're gonna dig into that a little bit later. Let's let's go back over to you. Explain adaptive data governance. Help us understand that >>really adaptive data governance is about achieving business outcomes through automation. It's really also about establishing a data driven culture and pushing what's traditionally managed in I t out to the business. And to do that, you've got to you've got Thio. You've got to enable an environment where people can actually access and look at the information about the data, not necessarily access the underlying data because we've got privacy concerns itself. But they need to understand what kind of data they have, what shape it's in what's dependent on it upstream and downstream, and so that they could make their educated decisions on on what they need to do to achieve those business outcomes. >>Ah, >>lot of a lot of frameworks these days are hardwired, so you can set up a set of business rules, and that set of business rules works for a very specific database and a specific schema. But imagine a world where you could just >>say, you >>know, the start date of alone must always be before the end date of alone and having that generic rule, regardless of the underlying database and applying it even when a new database comes online and having those rules applied. That's what adaptive data governance about I like to think of. It is the intersection of three circles, Really. It's the technical metadata coming together with policies and rules and coming together with the business ontology ease that are that are unique to that particular business. And this all of this. Bringing this all together allows you to enable rapid change in your environment. So it's a mouthful, adaptive data governance. But that's what it kind of comes down to. >>So, Angie, help me understand this. Is this book enterprise companies are doing now? Are they not quite there yet. >>Well, you know, Lisa, I think every organization is is going at its pace. But, you know, markets are changing the economy and the speed at which, um, some of the changes in the economy happening is is compelling more businesses to look at being more digital in how they serve their own customers. Eh? So what we're seeing is a number of trends here from heads of data Chief Data Officers, CEO, stepping back from, ah, one size fits all approach because they've tried that before, and it it just hasn't worked. They've spent millions of dollars on I T programs China Dr Value from that data on Bennett. And they've ended up with large teams of manual processing around data to try and hardwire these policies to fit with the context and each line of business and on that hasn't worked. So the trends that we're seeing emerge really relate. Thio, How do I There's a chief data officer as a CEO. Inject more automation into a lot of these common tax. Andi, you know, we've been able toc that impact. I think the news here is you know, if you're trying to create a knowledge graph a data catalog or Ah, business glossary. And you're trying to do that manually will stop you. You don't have to do that manually anymore. I think best example I can give is Lester and I We we like Chinese food and Japanese food on. If you were sitting there with your chopsticks, you wouldn't eat the bowl of rice with the chopsticks, one grain at a time. What you'd want to do is to find a more productive way to to enjoy that meal before it gets cold. Andi, that's similar to how we're able to help the organizations to digest their data is to get through it faster, enjoy the benefits of putting that data to work. >>And if it was me eating that food with you guys, I would be not using chopsticks. I would be using a fork and probably a spoon. So eso Lester, how then does iota who go about doing this and enabling customers to achieve this? >>Let me, uh, let me show you a little story have here. So if you take a look at the challenges the most customers have, they're very similar, but every customers on a different data journey, so but it all starts with what data do I have? What questions or what shape is that data in? Uh, how is it structured? What's dependent on it? Upstream and downstream. Um, what insights can I derive from that data? And how can I answer all of those questions automatically? So if you look at the challenges for these data professionals, you know, they're either on a journey to the cloud. Maybe they're doing a migration oracle. Maybe they're doing some data governance changes on bits about enabling this. So if you look at these challenges and I'm gonna take you through a >>story here, E, >>I want to introduce Amanda. Man does not live like, uh, anyone in any large organization. She's looking around and she just sees stacks of data. I mean, different databases, the one she knows about, the one she doesn't know about what should know about various different kinds of databases. And a man is just tasking with understanding all of this so that they can embark on her data journey program. So So a man who goes through and she's great. I've got some handy tools. I can start looking at these databases and getting an idea of what we've got. Well, as she digs into the databases, she starts to see that not everything is as clear as she might have hoped it would be. You know, property names or column names, or have ambiguous names like Attribute one and attribute to or maybe date one and date to s Oh, man is starting to struggle, even though she's get tools to visualize. And look what look at these databases. She still No, she's got a long road ahead. And with 2000 databases in her large enterprise, yes, it's gonna be a long turkey but Amanda Smart. So she pulls out her trusty spreadsheet to track all of her findings on what she doesn't know about. She raises a ticket or maybe tries to track down the owner to find what the data means. And she's tracking all this information. Clearly, this doesn't scale that well for Amanda, you know? So maybe organization will get 10 Amanda's to sort of divide and conquer that work. But even that doesn't work that well because they're still ambiguities in the data with Iota ho. What we do is we actually profile the underlying data. By looking at the underlying data, we can quickly see that attribute. One looks very much like a U. S. Social Security number and attribute to looks like a I c D 10 medical code. And we do this by using anthologies and dictionaries and algorithms to help identify the underlying data and then tag it. Key Thio Doing, uh, this automation is really being able to normalize things across different databases, so that where there's differences in column names, I know that in fact, they contain contain the same data. And by going through this exercise with a Tahoe, not only can we identify the data, but we also could gain insights about the data. So, for example, we can see that 97% of that time that column named Attribute one that's got us Social Security numbers has something that looks like a Social Security number. But 3% of the time, it doesn't quite look right. Maybe there's a dash missing. Maybe there's a digit dropped. Or maybe there's even characters embedded in it. So there may be that may be indicative of a data quality issues, so we try to find those kind of things going a step further. We also try to identify data quality relationships. So, for example, we have two columns, one date, one date to through Ah, observation. We can see that date 1 99% of the time is less than date, too. 1% of the time. It's not probably indicative of a data quality issue, but going a step further, we can also build a business rule that says Day one is less than date to. And so then when it pops up again, we can quickly identify and re mediate that problem. So these are the kinds of things that we could do with with iota going even a step further. You could take your your favorite data science solution production ISAT and incorporated into our next version a zey what we call a worker process to do your own bespoke analytics. >>We spoke analytics. Excellent, Lester. Thank you. So a J talk us through some examples of where you're putting this to use. And also what is some of the feedback from >>some customers? But I think it helped do this Bring it to life a little bit. Lisa is just to talk through a case study way. Pull something together. I know it's available for download, but in ah, well known telecommunications media company, they had a lot of the issues that lasted. You spoke about lots of teams of Amanda's, um, super bright data practitioners, um, on baby looking to to get more productivity out of their day on, deliver a good result for their own customers for cell phone subscribers, Um, on broadband users. So you know that some of the examples that we can see here is how we went about auto generating a lot of that understanding off that data within hours. So Amanda had her data catalog populated automatically. A business class three built up on it. Really? Then start to see. Okay, where do I want Thio? Apply some policies to the data to to set in place some controls where they want to adapt, how different lines of business, maybe tax versus customer operations have different access or permissions to that data on What we've been able to do there is, is to build up that picture to see how does data move across the entire organization across the state. Andi on monitor that overtime for improvement, so have taken it from being a reactive. Let's do something Thio. Fix something. Thio, Now more proactive. We can see what's happening with our data. Who's using it? Who's accessing it, how it's being used, how it's being combined. Um, on from there. Taking a proactive approach is a real smart use of of the talents in in that telco organization Onda folks that worked there with data. >>Okay, Jason, dig into that a little bit deeper. And one of the things I was thinking when you were talking through some of those outcomes that you're helping customers achieve is our ally. How do customers measure are? Why? What are they seeing with iota host >>solution? Yeah, right now that the big ticket item is time to value on. And I think in data, a lot of the upfront investment cause quite expensive. They have been today with a lot of the larger vendors and technologies. So what a CEO and economic bio really needs to be certain of is how quickly can I get that are away. I think we've got something we can show. Just pull up a before and after, and it really comes down to hours, days and weeks. Um, where we've been able Thio have that impact on in this playbook that we pulled together before and after picture really shows. You know, those savings that committed a bit through providing data into some actionable form within hours and days to to drive agility, but at the same time being out and forced the controls to protect the use of that data who has access to it. So these are the number one thing I'd have to say. It's time on. We can see that on the the graphic that we've just pulled up here. >>We talk about achieving adaptive data governance. Lester, you guys talk about automation. You talk about machine learning. How are you seeing those technologies being a facilitator of organizations adopting adaptive data governance? Well, >>Azaz, we see Mitt Emmanuel day. The days of manual effort are so I think you know this >>is a >>multi step process. But the very first step is understanding what you have in normalizing that across your data estate. So you couple this with the ontology, that air unique to your business. There is no algorithms, and you basically go across and you identify and tag tag that data that allows for the next steps toe happen. So now I can write business rules not in terms of columns named columns, but I could write him in terms of the tags being able to automate. That is a huge time saver and the fact that we can suggest that as a rule, rather than waiting for a person to come along and say, Oh, wow. Okay, I need this rule. I need this will thes air steps that increased that are, I should say, decrease that time to value that A. J talked about and then, lastly, a couple of machine learning because even with even with great automation and being able to profile all of your data and getting a good understanding, that brings you to a certain point. But there's still ambiguities in the data. So, for example, I might have to columns date one and date to. I may have even observed the date. One should be less than day two, but I don't really know what date one and date to our other than a date. So this is where it comes in, and I might ask the user said, >>Can >>you help me identify what date? One and date You are in this in this table. Turns out they're a start date and an end date for alone That gets remembered, cycled into the machine learning. So if I start to see this pattern of date one day to elsewhere, I'm going to say, Is it start dating and date? And these Bringing all these things together with this all this automation is really what's key to enabling this This'll data governance. Yeah, >>great. Thanks. Lester and a j wanna wrap things up with something that you mentioned in the beginning about what you guys were doing with Oracle. Take us out by telling us what you're doing there. How are you guys working together? >>Yeah, I think those of us who worked in i t for many years we've We've learned Thio trust articles technology that they're shifting now to ah, hybrid on Prohm Cloud Generation to platform, which is exciting. Andi on their existing customers and new customers moving to article on a journey. So? So Oracle came to us and said, you know, we can see how quickly you're able to help us change mindsets Ondas mindsets are locked in a way of thinking around operating models of I t. That there may be no agile and what siloed on day wanting to break free of that and adopt a more agile A p I at driven approach. A lot of the work that we're doing with our recall no is around, uh, accelerating what customers conduce with understanding their data and to build digital APS by identifying the the underlying data that has value. Onda at the time were able to do that in in in hours, days and weeks. Rather many months. Is opening up the eyes to Chief Data Officers CEO to say, Well, maybe we can do this whole digital transformation this year. Maybe we can bring that forward and and transform who we are as a company on that's driving innovation, which we're excited about it. I know Oracle, a keen Thio to drive through and >>helping businesses transformed digitally is so incredibly important in this time as we look Thio things changing in 2021 a. J. Lester thank you so much for joining me on this segment explaining adaptive data governance, how organizations can use it benefit from it and achieve our Oi. Thanks so much, guys. >>Thank you. Thanks again, Lisa. >>In a moment, we'll look a adaptive data governance in banking. This is the Cube, your global leader in high tech coverage. >>Innovation, impact influence. Welcome to the Cube. Disruptors. Developers and practitioners learn from the voices of leaders who share their personal insights from the hottest digital events around the globe. Enjoy the best this community has to offer on the Cube, your global leader in high tech digital coverage. >>Our next segment here is an interesting panel you're gonna hear from three gentlemen about adaptive data. Governments want to talk a lot about that. Please welcome Yusuf Khan, the global director of data services for Iot Tahoe. We also have Santiago Castor, the chief data officer at the First Bank of Nigeria, and good John Vander Wal, Oracle's senior manager of digital transformation and industries. Gentlemen, it's great to have you joining us in this in this panel. Great >>to be >>tried for me. >>Alright, Santiago, we're going to start with you. Can you talk to the audience a little bit about the first Bank of Nigeria and its scale? This is beyond Nigeria. Talk to us about that. >>Yes, eso First Bank of Nigeria was created 125 years ago. One of the oldest ignored the old in Africa because of the history he grew everywhere in the region on beyond the region. I am calling based in London, where it's kind of the headquarters and it really promotes trade, finance, institutional banking, corporate banking, private banking around the world in particular, in relationship to Africa. We are also in Asia in in the Middle East. >>So, Sanjay, go talk to me about what adaptive data governance means to you. And how does it help the first Bank of Nigeria to be able to innovate faster with the data that you have? >>Yes, I like that concept off adaptive data governor, because it's kind of Ah, I would say an approach that can really happen today with the new technologies before it was much more difficult to implement. So just to give you a little bit of context, I I used to work in consulting for 16, 17 years before joining the president of Nigeria, and I saw many organizations trying to apply different type of approaches in the governance on by the beginning early days was really kind of a year. A Chicago A. A top down approach where data governance was seeing as implement a set of rules, policies and procedures. But really, from the top down on is important. It's important to have the battle off your sea level of your of your director. Whatever I saw, just the way it fails, you really need to have a complimentary approach. You can say bottom are actually as a CEO are really trying to decentralize the governor's. Really, Instead of imposing a framework that some people in the business don't understand or don't care about it, it really needs to come from them. So what I'm trying to say is that data basically support business objectives on what you need to do is every business area needs information on the detector decisions toe actually be able to be more efficient or create value etcetera. Now, depending on the business questions they have to solve, they will need certain data set. So they need actually to be ableto have data quality for their own. For us now, when they understand that they become the stores naturally on their own data sets. And that is where my bottom line is meeting my top down. You can guide them from the top, but they need themselves to be also empower and be actually, in a way flexible to adapt the different questions that they have in orderto be able to respond to the business needs. Now I cannot impose at the finish for everyone. I need them to adapt and to bring their answers toe their own business questions. That is adaptive data governor and all That is possible because we have. And I was saying at the very beginning just to finalize the point, we have new technologies that allow you to do this method data classifications, uh, in a very sophisticated way that you can actually create analitico of your metadata. You can understand your different data sources in order to be able to create those classifications like nationalities, a way of classifying your customers, your products, etcetera. >>So one of the things that you just said Santa kind of struck me to enable the users to be adaptive. They probably don't want to be logging in support ticket. So how do you support that sort of self service to meet the demand of the users so that they can be adaptive. >>More and more business users wants autonomy, and they want to basically be ableto grab the data and answer their own question. Now when you have, that is great, because then you have demand of businesses asking for data. They're asking for the insight. Eso How do you actually support that? I would say there is a changing culture that is happening more and more. I would say even the current pandemic has helped a lot into that because you have had, in a way, off course, technology is one of the biggest winners without technology. We couldn't have been working remotely without these technologies where people can actually looking from their homes and still have a market data marketplaces where they self serve their their information. But even beyond that data is a big winner. Data because the pandemic has shown us that crisis happened, that we cannot predict everything and that we are actually facing a new kind of situation out of our comfort zone, where we need to explore that we need to adapt and we need to be flexible. How do we do that with data. Every single company either saw the revenue going down or the revenue going very up For those companies that are very digital already. Now it changed the reality, so they needed to adapt. But for that they needed information. In order to think on innovate, try toe, create responses So that type of, uh, self service off data Haider for data in order to be able to understand what's happening when the prospect is changing is something that is becoming more, uh, the topic today because off the condemning because of the new abilities, the technologies that allow that and then you then are allowed to basically help your data. Citizens that call them in the organization people that no other business and can actually start playing and an answer their own questions. Eso so these technologies that gives more accessibility to the data that is some cataloging so they can understand where to go or what to find lineage and relationships. All this is is basically the new type of platforms and tools that allow you to create what are called a data marketplace. I think these new tools are really strong because they are now allowing for people that are not technology or I t people to be able to play with data because it comes in the digital world There. Used to a given example without your who You have a very interesting search functionality. Where if you want to find your data you want to sell, Sir, you go there in that search and you actually go on book for your data. Everybody knows how to search in Google, everybody's searching Internet. So this is part of the data culture, the digital culture. They know how to use those schools. Now, similarly, that data marketplace is, uh, in you can, for example, see which data sources they're mostly used >>and enabling that speed that we're all demanding today during these unprecedented times. Goodwin, I wanted to go to you as we talk about in the spirit of evolution, technology is changing. Talk to us a little bit about Oracle Digital. What are you guys doing there? >>Yeah, Thank you. Um, well, Oracle Digital is a business unit that Oracle EMEA on. We focus on emerging countries as well as low and enterprises in the mid market, in more developed countries and four years ago. This started with the idea to engage digital with our customers. Fear Central helps across EMEA. That means engaging with video, having conference calls, having a wall, a green wall where we stand in front and engage with our customers. No one at that time could have foreseen how this is the situation today, and this helps us to engage with our customers in the way we were already doing and then about my team. The focus of my team is to have early stage conversations with our with our customers on digital transformation and innovation. And we also have a team off industry experts who engaged with our customers and share expertise across EMEA, and we inspire our customers. The outcome of these conversations for Oracle is a deep understanding of our customer needs, which is very important so we can help the customer and for the customer means that we will help them with our technology and our resource is to achieve their goals. >>It's all about outcomes, right? Good Ron. So in terms of automation, what are some of the things Oracle's doing there to help your clients leverage automation to improve agility? So that they can innovate faster, which in these interesting times it's demanded. >>Yeah, thank you. Well, traditionally, Oracle is known for their databases, which have bean innovated year over year. So here's the first lunch on the latest innovation is the autonomous database and autonomous data warehouse. For our customers, this means a reduction in operational costs by 90% with a multi medal converts, database and machine learning based automation for full life cycle management. Our databases self driving. This means we automate database provisioning, tuning and scaling. The database is self securing. This means ultimate data protection and security, and it's self repairing the automates failure, detection fail over and repair. And then the question is for our customers, What does it mean? It means they can focus on their on their business instead off maintaining their infrastructure and their operations. >>That's absolutely critical use if I want to go over to you now. Some of the things that we've talked about, just the massive progression and technology, the evolution of that. But we know that whether we're talking about beta management or digital transformation, a one size fits all approach doesn't work to address the challenges that the business has, um that the i t folks have, as you're looking through the industry with what Santiago told us about first Bank of Nigeria. What are some of the changes that you're seeing that I owe Tahoe seeing throughout the industry? >>Uh, well, Lisa, I think the first way I'd characterize it is to say, the traditional kind of top down approach to data where you have almost a data Policeman who tells you what you can and can't do, just doesn't work anymore. It's too slow. It's too resource intensive. Uh, data management data, governments, digital transformation itself. It has to be collaborative on. There has to be in a personalization to data users. Um, in the environment we find ourselves in. Now, it has to be about enabling self service as well. Um, a one size fits all model when it comes to those things around. Data doesn't work. As Santiago was saying, it needs to be adapted toe how the data is used. Andi, who is using it on in order to do this cos enterprises organizations really need to know their data. They need to understand what data they hold, where it is on what the sensitivity of it is they can then any more agile way apply appropriate controls on access so that people themselves are and groups within businesses are our job and could innovate. Otherwise, everything grinds to a halt, and you risk falling behind your competitors. >>Yeah, that one size fits all term just doesn't apply when you're talking about adaptive and agility. So we heard from Santiago about some of the impact that they're making with First Bank of Nigeria. Used to talk to us about some of the business outcomes that you're seeing other customers make leveraging automation that they could not do >>before it's it's automatically being able to classify terabytes, terabytes of data or even petabytes of data across different sources to find duplicates, which you can then re mediate on. Deletes now, with the capabilities that iota offers on the Oracle offers, you can do things not just where the five times or 10 times improvement, but it actually enables you to do projects for Stop that otherwise would fail or you would just not be able to dio I mean, uh, classifying multi terrible and multi petabytes states across different sources, formats very large volumes of data in many scenarios. You just can't do that manually. I mean, we've worked with government departments on the issues there is expect are the result of fragmented data. There's a lot of different sources. There's lot of different formats and without these newer technologies to address it with automation on machine learning, the project isn't durable. But now it is on that that could lead to a revolution in some of these businesses organizations >>to enable that revolution that there's got to be the right cultural mindset. And one of the when Santiago was talking about folks really kind of adapted that. The thing I always call that getting comfortably uncomfortable. But that's hard for organizations to. The technology is here to enable that. But well, you're talking with customers use. How do you help them build the trust in the confidence that the new technologies and a new approaches can deliver what they need? How do you help drive the kind of a tech in the culture? >>It's really good question is because it can be quite scary. I think the first thing we'd start with is to say, Look, the technology is here with businesses like I Tahoe. Unlike Oracle, it's already arrived. What you need to be comfortable doing is experimenting being agile around it, Andi trying new ways of doing things. Uh, if you don't wanna get less behind that Santiago on the team that fbn are a great example off embracing it, testing it on a small scale on, then scaling up a Toyota, we offer what we call a data health check, which can actually be done very quickly in a matter of a few weeks. So we'll work with a customer. Picky use case, install the application, uh, analyzed data. Drive out Cem Cem quick winds. So we worked in the last few weeks of a large entity energy supplier, and in about 20 days, we were able to give them an accurate understanding of their critical data. Elements apply. Helping apply data protection policies. Minimize copies of the data on work out what data they needed to delete to reduce their infrastructure. Spend eso. It's about experimenting on that small scale, being agile on, then scaling up in a kind of very modern way. >>Great advice. Uh, Santiago, I'd like to go back to Is we kind of look at again that that topic of culture and the need to get that mindset there to facilitate these rapid changes, I want to understand kind of last question for you about how you're doing that from a digital transformation perspective. We know everything is accelerating in 2020. So how are you building resilience into your data architecture and also driving that cultural change that can help everyone in this shift to remote working and a lot of the the digital challenges and changes that we're all going through? >>The new technologies allowed us to discover the dating anyway. Toe flawed and see very quickly Information toe. Have new models off over in the data on giving autonomy to our different data units. Now, from that autonomy, they can then compose an innovator own ways. So for me now, we're talking about resilience because in a way, autonomy and flexibility in a organization in a data structure with platform gives you resilience. The organizations and the business units that I have experienced in the pandemic are working well. Are those that actually because they're not physically present during more in the office, you need to give them their autonomy and let them actually engaged on their own side that do their own job and trust them in a way on as you give them, that they start innovating and they start having a really interesting ideas. So autonomy and flexibility. I think this is a key component off the new infrastructure. But even the new reality that on then it show us that, yes, we used to be very kind off structure, policies, procedures as very important. But now we learn flexibility and adaptability of the same side. Now, when you have that a key, other components of resiliency speed, because people want, you know, to access the data and access it fast and on the site fast, especially changes are changing so quickly nowadays that you need to be ableto do you know, interact. Reiterate with your information to answer your questions. Pretty, um, so technology that allows you toe be flexible iterating on in a very fast job way continue will allow you toe actually be resilient in that way, because you are flexible, you adapt your job and you continue answering questions as they come without having everything, setting a structure that is too hard. We also are a partner off Oracle and Oracle. Embodies is great. They have embedded within the transactional system many algorithms that are allowing us to calculate as the transactions happened. What happened there is that when our customers engaged with algorithms and again without your powers, well, the machine learning that is there for for speeding the automation of how you find your data allows you to create a new alliance with the machine. The machine is their toe, actually, in a way to your best friend to actually have more volume of data calculated faster. In a way, it's cover more variety. I mean, we couldn't hope without being connected to this algorithm on >>that engagement is absolutely critical. Santiago. Thank you for sharing that. I do wanna rap really quickly. Good On one last question for you, Santiago talked about Oracle. You've talked about a little bit. As we look at digital resilience, talk to us a little bit in the last minute about the evolution of Oracle. What you guys were doing there to help your customers get the resilience that they have toe have to be not just survive but thrive. >>Yeah. Oracle has a cloud offering for infrastructure, database, platform service and a complete solutions offered a South on Daz. As Santiago also mentioned, We are using AI across our entire portfolio and by this will help our customers to focus on their business innovation and capitalize on data by enabling new business models. Um, and Oracle has a global conference with our cloud regions. It's massively investing and innovating and expanding their clouds. And by offering clouds as public cloud in our data centers and also as private cloud with clouded customer, we can meet every sovereignty and security requirements. And in this way we help people to see data in new ways. We discover insights and unlock endless possibilities. And and maybe 11 of my takeaways is if I If I speak with customers, I always tell them you better start collecting your data. Now we enable this partners like Iota help us as well. If you collect your data now, you are ready for tomorrow. You can never collect your data backwards, So that is my take away for today. >>You can't collect your data backwards. Excellently, John. Gentlemen, thank you for sharing all of your insights. Very informative conversation in a moment, we'll address the question. Do you know your data? >>Are you interested in test driving the iota Ho platform kick Start the benefits of data automation for your business through the Iota Ho Data Health check program. Ah, flexible, scalable sandbox environment on the cloud of your choice with set up service and support provided by Iota ho. Look time with a data engineer to learn more and see Io Tahoe in action from around the globe. It's the Cube presenting adaptive data governance brought to you by Iota Ho. >>In this next segment, we're gonna be talking to you about getting to know your data. And specifically you're gonna hear from two folks at Io Tahoe. We've got enterprise account execs to be to Davis here, as well as Enterprise Data engineer Patrick Simon. They're gonna be sharing insights and tips and tricks for how you could get to know your data and quickly on. We also want to encourage you to engage with the media and Patrick, use the chat feature to the right, send comments, questions or feedback so you can participate. All right, Patrick Savita, take it away. Alright. >>Thankfully saw great to be here as Lisa mentioned guys, I'm the enterprise account executive here in Ohio. Tahoe you Pat? >>Yeah. Hey, everyone so great to be here. I said my name is Patrick Samit. I'm the enterprise data engineer here in Ohio Tahoe. And we're so excited to be here and talk about this topic as one thing we're really trying to perpetuate is that data is everyone's business. >>So, guys, what patent I got? I've actually had multiple discussions with clients from different organizations with different roles. So we spoke with both your technical and your non technical audience. So while they were interested in different aspects of our platform, we found that what they had in common was they wanted to make data easy to understand and usable. So that comes back. The pats point off to being everybody's business because no matter your role, we're all dependent on data. So what Pan I wanted to do today was wanted to walk you guys through some of those client questions, slash pain points that we're hearing from different industries and different rules and demo how our platform here, like Tahoe, is used for automating Dozier related tasks. So with that said are you ready for the first one, Pat? >>Yeah, Let's do it. >>Great. So I'm gonna put my technical hat on for this one. So I'm a data practitioner. I just started my job. ABC Bank. I have, like, over 100 different data sources. So I have data kept in Data Lakes, legacy data, sources, even the cloud. So my issue is I don't know what those data sources hold. I don't know what data sensitive, and I don't even understand how that data is connected. So how can I saw who help? >>Yeah, I think that's a very common experience many are facing and definitely something I've encountered in my past. Typically, the first step is to catalog the data and then start mapping the relationships between your various data stores. Now, more often than not, this has tackled through numerous meetings and a combination of excel and something similar to video which are too great tools in their own part. But they're very difficult to maintain. Just due to the rate that we are creating data in the modern world. It starts to beg for an idea that can scale with your business needs. And this is where a platform like Io Tahoe becomes so appealing, you can see here visualization of the data relationships created by the I. O. Tahoe service. Now, what is fantastic about this is it's not only laid out in a very human and digestible format in the same action of creating this view, the data catalog was constructed. >>Um so is the data catalog automatically populated? Correct. Okay, so So what I'm using Iota hope at what I'm getting is this complete, unified automated platform without the added cost? Of course. >>Exactly. And that's at the heart of Iota Ho. A great feature with that data catalog is that Iota Ho will also profile your data as it creates the catalog, assigning some meaning to those pesky column underscore ones and custom variable underscore tents. They're always such a joy to deal with. Now, by leveraging this interface, we can start to answer the first part of your question and understand where the core relationships within our data exists. Uh, personally, I'm a big fan of this view, as it really just helps the i b naturally John to these focal points that coincide with these key columns following that train of thought, Let's examine the customer I D column that seems to be at the center of a lot of these relationships. We can see that it's a fairly important column as it's maintaining the relationship between at least three other tables. >>Now you >>notice all the connectors are in this blue color. This means that their system defined relationships. But I hope Tahoe goes that extra mile and actually creates thes orange colored connectors as well. These air ones that are machine learning algorithms have predicted to be relationships on. You can leverage to try and make new and powerful relationships within your data. >>Eso So this is really cool, and I can see how this could be leverage quickly now. What if I added new data sources or your multiple data sources and need toe identify what data sensitive can iota who detect that? >>Yeah, definitely. Within the hotel platform. There, already over 300 pre defined policies such as hip for C, C, P. A and the like one can choose which of these policies to run against their data along for flexibility and efficiency and running the policies that affect organization. >>Okay, so so 300 is an exceptional number. I'll give you that. But what about internal policies that apply to my organization? Is there any ability for me to write custom policies? >>Yeah, that's no issue. And it's something that clients leverage fairly often to utilize this function when simply has to write a rejects that our team has helped many deploy. After that, the custom policy is stored for future use to profile sensitive data. One then selects the data sources they're interested in and select the policies that meet your particular needs. The interface will automatically take your data according to the policies of detects, after which you can review the discoveries confirming or rejecting the tagging. All of these insights are easily exported through the interface. Someone can work these into the action items within your project management systems, and I think this lends to the collaboration as a team can work through the discovery simultaneously, and as each item is confirmed or rejected, they can see it ni instantaneously. All this translates to a confidence that with iota hope, you can be sure you're in compliance. >>So I'm glad you mentioned compliance because that's extremely important to my organization. So what you're saying when I use the eye a Tahoe automated platform, we'd be 90% more compliant that before were other than if you were going to be using a human. >>Yeah, definitely the collaboration and documentation that the Iot Tahoe interface lends itself to really help you build that confidence that your compliance is sound. >>So we're planning a migration. Andi, I have a set of reports I need to migrate. But what I need to know is, uh well, what what data sources? Those report those reports are dependent on. And what's feeding those tables? >>Yeah, it's a fantastic questions to be toe identifying critical data elements, and the interdependencies within the various databases could be a time consuming but vital process and the migration initiative. Luckily, Iota Ho does have an answer, and again, it's presented in a very visual format. >>Eso So what I'm looking at here is my entire day landscape. >>Yes, exactly. >>Let's say I add another data source. I can still see that unified 3 60 view. >>Yeah, One future that is particularly helpful is the ability to add data sources after the data lineage. Discovery has finished alone for the flexibility and scope necessary for any data migration project. If you only need need to select a few databases or your entirety, this service will provide the answers. You're looking for things. Visual representation of the connectivity makes the identification of critical data elements a simple matter. The connections air driven by both system defined flows as well as those predicted by our algorithms, the confidence of which, uh, can actually be customized to make sure that they're meeting the needs of the initiative that you have in place. This also provides tabular output in case you needed for your own internal documentation or for your action items, which we can see right here. Uh, in this interface, you can actually also confirm or deny the pair rejection the pair directions, allowing to make sure that the data is as accurate as possible. Does that help with your data lineage needs? >>Definitely. So So, Pat, My next big question here is So now I know a little bit about my data. How do I know I can trust >>it? So >>what I'm interested in knowing, really is is it in a fit state for me to use it? Is it accurate? Does it conform to the right format? >>Yeah, that's a great question. And I think that is a pain point felt across the board, be it by data practitioners or data consumers alike. Another service that I owe Tahoe provides is the ability to write custom data quality rules and understand how well the data pertains to these rules. This dashboard gives a unified view of the strength of these rules, and your dad is overall quality. >>Okay, so Pat s o on on the accuracy scores there. So if my marketing team needs to run, a campaign can read dependent those accuracy scores to know what what tables have quality data to use for our marketing campaign. >>Yeah, this view would allow you to understand your overall accuracy as well as dive into the minutia to see which data elements are of the highest quality. So for that marketing campaign, if you need everything in a strong form, you'll be able to see very quickly with these high level numbers. But if you're only dependent on a few columns to get that information out the door, you can find that within this view, eso >>you >>no longer have to rely on reports about reports, but instead just come to this one platform to help drive conversations between stakeholders and data practitioners. >>So I get now the value of IATA who brings by automatically capturing all those technical metadata from sources. But how do we match that with the business glossary? >>Yeah, within the same data quality service that we just reviewed, one can actually add business rules detailing the definitions and the business domains that these fall into. What's more is that the data quality rules were just looking at can then be tied into these definitions. Allowing insight into the strength of these business rules is this service that empowers stakeholders across the business to be involved with the data life cycle and take ownership over the rules that fall within their domain. >>Okay, >>so those custom rules can I apply that across data sources? >>Yeah, you could bring in as many data sources as you need, so long as you could tie them to that unified definition. >>Okay, great. Thanks so much bad. And we just want to quickly say to everyone working in data, we understand your pain, so please feel free to reach out to us. we are Website the chapel. Oh, Arlington. And let's get a conversation started on how iota Who can help you guys automate all those manual task to help save you time and money. Thank you. Thank >>you. Your Honor, >>if I could ask you one quick question, how do you advise customers? You just walk in this great example this banking example that you instantly to talk through. How do you advise customers get started? >>Yeah, I think the number one thing that customers could do to get started with our platform is to just run the tag discovery and build up that data catalog. It lends itself very quickly to the other needs you might have, such as thes quality rules. A swell is identifying those kind of tricky columns that might exist in your data. Those custom variable underscore tens I mentioned before >>last questions to be to anything to add to what Pat just described as a starting place. >>I'm no, I think actually passed something that pretty well, I mean, just just by automating all those manual task. I mean, it definitely can save your company a lot of time and money, so we we encourage you just reach out to us. Let's get that conversation >>started. Excellent. So, Pete and Pat, thank you so much. We hope you have learned a lot from these folks about how to get to know your data. Make sure that it's quality, something you can maximize the value of it. Thanks >>for watching. Thanks again, Lisa, for that very insightful and useful deep dive into the world of adaptive data governance with Iota Ho Oracle First Bank of Nigeria This is Dave a lot You won't wanna mess Iota, whose fifth episode in the data automation Siri's in that we'll talk to experts from Red Hat and Happiest Minds about their best practices for managing data across hybrid cloud Inter Cloud multi Cloud I T environment So market calendar for Wednesday, January 27th That's Episode five. You're watching the Cube Global Leader digital event technique

>>Thank you. Thank you so much for having me. I'm thrilled to be in conversation with you today. And I thought I would just kick things off with some opening reflections on this really important session theme, and then we can jump into discussion. So I'd like us to, as a starting point, um, wrestle with these buzz words, empowerment and inclusion so that we can, um, have them be more than kind of big platitudes and really have them reflected in our workplace cultures and the things that we design and the technologies that we put out into the world. And so to do that, I think we have to move beyond techno determinism and I'll explain what that means in just a minute. And techno determinism comes in two forms. The first on your left is the idea that technology automate. Um, all of these emerging trends are going to harm us are going to necessarily, um, harm humanity. >>They're going to take all the jobs they're going to remove human agency. This is what we might call the techno dystopian version of the story. And this is what Hollywood loves to sell us in the form of movies like the matrix or Terminator. The other version on your right is the techno utopian story that technologies automation, the robots, as a shorthand are going to save humanity. They're going to make everything more efficient, more equitable. And in this case, on the surface, they seem like opposing narratives, right? They're telling us different stories. At least they have different endpoints, but when you pull back the screen and look a little bit more closely, you see that they share an underlying logic, that technology is in the driver's seat and that human beings, that social society can just respond to what's happening. But we don't, I really have a say in what technologies are designed. >>And so to move beyond techno determinism, the notion that technology is in the driver's seat, we have to put the human agents and agencies back into the story protagonists and think carefully about what the human desires, worldviews values assumptions are that animate the production of technology. We have to put the humans behind the screen back into view. And so that's a very first step in when we do that. We see as was already mentioned that it's a very homogenous group right now in terms of who gets the power and the resources to produce the digital and physical infrastructure that everyone else has to live with. And so, as a first step, we need to think about how to, to create more participation of those who are working behind the scenes to design technology. Now, to dig a little more deeper into this, I want to offer a kind of low tech example before we get to the more high tech ones. >>So what you see in front of you here is a simple park bench public it's located in Berkeley, California, which is where I went to graduate school. And on this one particular visit, I was living in Boston. And so I was back in California, it was February, it was freezing where I was coming from. And so I wanted to take a few minutes in between meetings to just lay out in the sun and soak in some vitamin D. And I quickly realized actually I couldn't lay down on the bench because of the way it had been designed with these arm rests at intermittent intervals. And so here I thought, okay, th th the armrests have a functional reason why they're there. I mean, you could literally rest your elbows there, or, um, you know, it can create a little bit of privacy of someone sitting there that you don't know. >>Um, when I was nine months pregnant, it could help me get up and down or for the elderly the same thing. So it has a lot of functional reasons, but I also thought about the fact that it prevents people who are, are homeless from sleeping on the bench. And this is the Bay area that we're talking about, where in fact, the tech boom has gone hand in hand with a housing crisis. Those things have grown in tandem. So innovation has grown with inequity because we have, I haven't thought carefully about how to address the social context in which technology grows and blossoms. And so I thought, okay, this crisis is growing in this area. And so perhaps this is a deliberate attempt to make sure that people don't sleep on the benches by the way that they're designed and where the, where they're implemented. And so this is what we might call structural inequity, by the way something is designed. >>It has certain yeah. Affects that exclude or harm different people. And so it may not necessarily be the intent, but that's the effect. And I did a little digging and I found, in fact, it's a global phenomenon, this thing that architect next call, hostile architecture around single occupancy, benches and Helsinki. So only one booty at a time, no Nolan down there. I've found caged benches in France. Yeah. And in this particular town, what's interesting here is that the mayor put these benches out in this little shopping Plaza and within 24 hours, the people in the town rally together and have them removed. So we see here that just because we, we have a discriminatory design in our public space, doesn't mean we have to live with it. We can actually work together to ensure that our public space reflects our better values. But I think my favorite example of all is the metered bench. >>And then this case, this bench is designed with spikes in them and to get the spikes to retreat into the bench, you have to feed the meter. You have to put some coins in, and I think it buys you about 15, 20 minutes, then the spikes come back up. And so you will be happy to know that in this case, uh, this was designed by a German artist to get people to think critically about issues of design, not the design of physical space, but the design of all kinds of things, public policies. And so we can think about how our public life in general is metered, that it serves those that can pay the price and others are excluded or harmed. Whether we're talking about education or healthcare. And the meter bench also presents something interesting for those of us who care about technology, it creates a technical fix for a social problem. >>In fact, it started out as art, but some municipalities in different parts of the world have actually adopted this in their public spaces, in their parks in order to deter so-called loiters from using that space. And so by a technical fix, we mean something that creates a short-term effect, right? It gets people who may want to sleep on it out of sight. They're unable to use it, but it doesn't address the underlying problems that create that need to sleep outside of the first place. And so, in addition to techno determinism, we have to think critically about technical fixes, that don't address the underlying issues that the tech tech technology is meant to solve. And so this is part of a broader issue of discriminatory design, and we can apply the bench metaphor to all kinds of things that we work with, or that we create. >>And the question we really have to continuously ask ourselves is what values are we building in to the physical and digital infrastructures around us? What are the spikes that we may unwittingly put into place? Or perhaps we didn't create the spikes. Perhaps we started a new job or a new position, and someone hands us something, this is the way things have always been done. So we inherit the spiked bench. What is our responsibility? When we notice that it's creating these kinds of harms or exclusions or technical fixes that are bypassing the underlying problem, what is our responsibility? All of this came to a head in the context of financial technologies. I don't know how many of you remember these high profile cases of tech insiders and CEOs who applied for apples, >>The Apple card. And in one case, a husband and wife applied, and the husband, the husband received a much higher limit, almost 20 times the limit as his, >>His wife, even though they shared bank accounts, they lived in common law state. Yeah. >>And so the question there was not only the fact that >>The husband was receiving a much better rate and a high and a better >>The interest rate and the limit, but also that there was no mechanism for the individuals involved to dispute what was happening. They didn't even know how, what the factors were that they were being judged that was creating this form of discrimination. So >>In terms of financial technologies, it's not simply the outcome, that's the issue, or that can be discriminatory, >>But the process that black box is all of the decision-making that makes it so that consumers and the general public have no way to question it, no way to understand how they're being judged adversely. And so it's the process, not only the product that we have to care a lot about. And so the case of the Apple card is part of a much broader phenomenon >>Of, um, races >>And sexist robots. This is how the headlines framed it a few years ago. And I was so interested in this framing because there was a first wave of stories that seemed to be shocked at the prospect, that technology is not neutral. Then there was a second wave of stories that seemed less surprised. Well, of course, technology inherits its creators biases. And now I think we've entered a phase of attempts to override and address the default settings of so-called racist and sexist robots for better or worse than here. Robots is just a kind of shorthand that the way that people are talking about automation and emerging technologies more broadly. And so, as I was encountering these headlines, I was thinking about how these are not problems simply brought on by machine learning or AI. They're not all brand new. And so I wanted to contribute to the conversation, a kind of larger context and a longer history for us to think carefully about the social dimensions of technology. And so I developed a concept called the new Jim code, >>Which plays on the phrase, >>Jim Crow, which is the way that the regime of white supremacy and inequality in this country was defined in a previous era. And I wanted us to think about how that legacy continues to haunt the present, how we might be coding bias into emerging technologies and the danger being that we imagine those technologies to be objective. And so this gives us a language to be able to name this phenomenon so that we can address it and change it under this larger umbrella of the new Jim code are four distinct ways that this phenomenon takes shape from the more obvious engineered inequity. Those are the kinds of inequalities tech mediated in the qualities that we can generally see coming. They're kind of obvious, but then we go down the line and we see it becomes harder to detect it's happening in our own backyards, it's happening around us. And we don't really have a view into the black box. And so it becomes more insidious. And so in the remaining couple of minutes, I'm just, just going to give you a taste of the last three of these, and then a move towards conclusion. Then we can start chatting. So when it comes to default discrimination, this is the way that social inequalities >>Become embedded in emerging technologies because designers of these technologies, aren't thinking carefully about history and sociology. A great example of this, uh, came to, um, uh, the headlines last fall when it was found that widely used healthcare algorithm, effecting millions of patients, um, was discriminating against black patients. And so what's especially important to note here is that this algorithm, healthcare algorithm does not explicitly take note of race. That is to say it is race neutral by using cost to predict healthcare needs this digital triaging system unwittingly reproduces health disparities, because on average, black people have incurred fewer costs for a variety of reasons, including structural inequality. So in my review of this study, by Obermeyer and colleagues, I want to draw attention to how indifference to social reality can be even more harmful than malicious intent. It doesn't have to be the intent of the designers to create this effect. >>And so we have to look carefully at how indifference is operating and how race neutrality can be a deadly force. When we move on to the next iteration of the new Jim code, coded exposure, there's a tension because on the one hand, you see this image where the darker skin individual is not being detected by the facial recognition system, right on the camera, on the computer. And so coded exposure names, this tension between wanting to be seen and included and recognized whether it's in facial recognition or in recommendation systems or in tailored advertising. But the opposite of that, the tension is with when you're over, it >>Included when you're surveilled, when you're >>Too centered. And so we should note that it's not simply in being left out, that's the problem, but it's in being included in harmful ways. And so I want us to think carefully about the rhetoric of inclusion and understand that inclusion is not simply an end point, it's a process, and it is possible to include people in harmful processes. And so we want to ensure that the process is not harmful for it to really be effective. The last iteration of the new Jim code. That means the, the most insidious let's say is technologies that are touted as helping us address bias. So they're not simply including people, but they're actively working to address bias. And so in this case, there are a lot of different companies that are using AI to hire, uh, create hiring, um, software and hiring algorithms, including this one higher view. >>And the idea is that there there's a lot that, um, AI can keep track of that human beings might miss. And so, so the software can make data-driven talent decisions after all the problem of employment discrimination is widespread and well-documented, so the logic goes, wouldn't this be even more reason to outsource decisions to AI? Well, let's think about this carefully. And this is the idea of techno benevolence, trying to do good without fully reckoning with what, how technology can reproduce inequalities. So some colleagues of mine at Princeton, um, tested a natural learning processing algorithm and was looking to see whether it exhibited the same, um, tendencies that psychologists have documented among humans. And what they found was that in fact, the algorithm associated black names with negative words and white names with pleasant sounding words. And so this particular audit builds on a classic study done around 2003 before all of the emerging technologies were on the scene where two university of Chicago economists sent out thousands of resumes to employers in Boston and Chicago. >>And all they did was change the names on those resumes. All of the other work history education were the same. And then they waited to see who would get called back and the applicants, the fictional applicants with white sounding names received 50% more callbacks than the, the black applicants. So if you're presented with that study, you might be tempted to say, well, let's let technology handle it since humans are so biased. But my colleagues here in computer science found that this natural language processing algorithm actually reproduced those same associations with black and white names. So two with gender coded words and names as Amazon learned a couple years ago, when its own hiring algorithm was found discriminating against women, nevertheless, it should be clear by now why technical fixes that claim to bypass human biases are so desirable. If only there was a way to slay centuries of racist and sexist demons with a social justice bot beyond desirable, more like magical, magical for employers, perhaps looking to streamline the grueling work of recruitment, but a curse from any job seekers as this headline puts it. >>Your next interview could be with a racist bot, bringing us back to that problem space. We started with just a few minutes ago. So it's worth noting that job seekers are already developing ways to subvert the system by trading answers to employers tests and creating fake applications as informal audits of their own. In terms of a more collective response. There's a Federation of European trade unions call you and I global that's developed a charter of digital rights for workers that touches on automated and AI based decisions to be included in bargaining agreements. And so this is one of many efforts to change the ecosystem, to change the context in which technology is being deployed to ensure more protections and more rights for everyday people in the U S there's the algorithmic accountability bill that's been presented. And it's one effort to create some more protections around this ubiquity of automated decisions. >>And I think we should all be calling for more public accountability when it comes to the widespread use of automated decisions. Another development that keeps me somewhat hopeful is that tech workers themselves are increasingly speaking out against the most egregious forms of corporate collusion with state sanctioned racism. And to get a taste of that, I encourage you to check out the hashtag tech, won't build it among other statements that they've made and walking out and petitioning their companies. One group said as the, at Google at Microsoft wrote as the people who build the technologies that Microsoft profits from, we refuse to be complicit in terms of education, which is my own ground zero. Um, it's a place where we can, we can grow a more historically and socially literate approach to tech design. And this is just one resource that you all can download, um, by developed by some wonderful colleagues at the data and society research Institute in New York. >>And the, the goal of this intervention is threefold to develop an intellectual understanding of how structural racism operates and algorithms, social media platforms and technologies not yet developed and emotional intelligence concerning how to resolve racially stressful situations within organizations and a commitment to take action, to reduce harms to communities of color. And so as a final way to think about why these things are so important, I want to offer, uh, a couple last provocations. The first is pressed to think a new about what actually is deep learning when it comes to computation. I want to suggest that computational depth when it comes to AI systems without historical or social depth is actually superficial learning. And so we need to have a much more interdisciplinary, integrated approach to knowledge production and to observing and understanding patterns that don't simply rely on one discipline in order to map reality. >>The last provocation is this. If as I suggested at the start in the inequity is woven into the very fabric of our society. It's built into the design of our, our policies, our physical infrastructures, and now even our digital infrastructures. That means that each twist coil and code is a chance for us to weave new patterns, practices, and politics. The vastness of the problems that we're up against will be their undoing. Once we, that we are pattern makers. So what does that look like? It looks like refusing colorblindness as an anecdote to tech media discrimination, rather than refusing to see difference. Let's take stock of how the training data and the models that we're creating. Have these built in decisions from the past that have often been discriminatory. It means actually thinking about the underside of inclusion, which can be targeting and how do we create a more participatory rather than predatory form of inclusion. And ultimately it also means owning our own power in these systems so that we can change the patterns of the past. If we're, if we inherit a spiked bench, that doesn't mean that we need to continue using it. We can work together to design more, just an equitable technologies. So with that, I look forward to our conversation.

>>don't talk mhm, >>Okay, thing is my presentation on coherent nonlinear dynamics and combinatorial optimization. This is going to be a talk to introduce an approach we're taking to the analysis of the performance of coherent using machines. So let me start with a brief introduction to easing optimization. The easing model represents a set of interacting magnetic moments or spins the total energy given by the expression shown at the bottom left of this slide. Here, the signal variables are meditate binary values. The Matrix element J. I. J. Represents the interaction, strength and signed between any pair of spins. I. J and A Chive represents a possible local magnetic field acting on each thing. The easing ground state problem is to find an assignment of binary spin values that achieves the lowest possible value of total energy. And an instance of the easing problem is specified by giving numerical values for the Matrix J in Vector H. Although the easy model originates in physics, we understand the ground state problem to correspond to what would be called quadratic binary optimization in the field of operations research and in fact, in terms of computational complexity theory, it could be established that the easing ground state problem is np complete. Qualitatively speaking, this makes the easing problem a representative sort of hard optimization problem, for which it is expected that the runtime required by any computational algorithm to find exact solutions should, as anatomically scale exponentially with the number of spends and for worst case instances at each end. Of course, there's no reason to believe that the problem instances that actually arrives in practical optimization scenarios are going to be worst case instances. And it's also not generally the case in practical optimization scenarios that we demand absolute optimum solutions. Usually we're more interested in just getting the best solution we can within an affordable cost, where costs may be measured in terms of time, service fees and or energy required for a computation. This focuses great interest on so called heuristic algorithms for the easing problem in other NP complete problems which generally get very good but not guaranteed optimum solutions and run much faster than algorithms that are designed to find absolute Optima. To get some feeling for present day numbers, we can consider the famous traveling salesman problem for which extensive compilations of benchmarking data may be found online. A recent study found that the best known TSP solver required median run times across the Library of Problem instances That scaled is a very steep route exponential for end up to approximately 4500. This gives some indication of the change in runtime scaling for generic as opposed the worst case problem instances. Some of the instances considered in this study were taken from a public library of T SPS derived from real world Veil aside design data. This feels I TSP Library includes instances within ranging from 131 to 744,710 instances from this library with end between 6880 13,584 were first solved just a few years ago in 2017 requiring days of run time and a 48 core to King hurts cluster, while instances with and greater than or equal to 14,233 remain unsolved exactly by any means. Approximate solutions, however, have been found by heuristic methods for all instances in the VLS i TSP library with, for example, a solution within 0.14% of a no lower bound, having been discovered, for instance, with an equal 19,289 requiring approximately two days of run time on a single core of 2.4 gigahertz. Now, if we simple mindedly extrapolate the root exponential scaling from the study up to an equal 4500, we might expect that an exact solver would require something more like a year of run time on the 48 core cluster used for the N equals 13,580 for instance, which shows how much a very small concession on the quality of the solution makes it possible to tackle much larger instances with much lower cost. At the extreme end, the largest TSP ever solved exactly has an equal 85,900. This is an instance derived from 19 eighties VLSI design, and it's required 136 CPU. Years of computation normalized to a single cord, 2.4 gigahertz. But the 24 larger so called world TSP benchmark instance within equals 1,904,711 has been solved approximately within ophthalmology. Gap bounded below 0.474%. Coming back to the general. Practical concerns have applied optimization. We may note that a recent meta study analyzed the performance of no fewer than 37 heuristic algorithms for Max cut and quadratic pioneer optimization problems and found the performance sort and found that different heuristics work best for different problem instances selected from a large scale heterogeneous test bed with some evidence but cryptic structure in terms of what types of problem instances were best solved by any given heuristic. Indeed, their their reasons to believe that these results from Mexico and quadratic binary optimization reflected general principle of performance complementarity among heuristic optimization algorithms in the practice of solving heart optimization problems there. The cerise is a critical pre processing issue of trying to guess which of a number of available good heuristic algorithms should be chosen to tackle a given problem. Instance, assuming that any one of them would incur high costs to run on a large problem, instances incidence, making an astute choice of heuristic is a crucial part of maximizing overall performance. Unfortunately, we still have very little conceptual insight about what makes a specific problem instance, good or bad for any given heuristic optimization algorithm. This has certainly been pinpointed by researchers in the field is a circumstance that must be addressed. So adding this all up, we see that a critical frontier for cutting edge academic research involves both the development of novel heuristic algorithms that deliver better performance, with lower cost on classes of problem instances that are underserved by existing approaches, as well as fundamental research to provide deep conceptual insight into what makes a given problem in, since easy or hard for such algorithms. In fact, these days, as we talk about the end of Moore's law and speculate about a so called second quantum revolution, it's natural to talk not only about novel algorithms for conventional CPUs but also about highly customized special purpose hardware architectures on which we may run entirely unconventional algorithms for combinatorial optimization such as easing problem. So against that backdrop, I'd like to use my remaining time to introduce our work on analysis of coherent using machine architectures and associate ID optimization algorithms. These machines, in general, are a novel class of information processing architectures for solving combinatorial optimization problems by embedding them in the dynamics of analog, physical or cyber physical systems, in contrast to both MAWR traditional engineering approaches that build using machines using conventional electron ICS and more radical proposals that would require large scale quantum entanglement. The emerging paradigm of coherent easing machines leverages coherent nonlinear dynamics in photonic or Opto electronic platforms to enable near term construction of large scale prototypes that leverage post Simoes information dynamics, the general structure of of current CM systems has shown in the figure on the right. The role of the easing spins is played by a train of optical pulses circulating around a fiber optical storage ring. A beam splitter inserted in the ring is used to periodically sample the amplitude of every optical pulse, and the measurement results are continually read into a refugee A, which uses them to compute perturbations to be applied to each pulse by a synchronized optical injections. These perturbations, air engineered to implement the spin, spin coupling and local magnetic field terms of the easing Hamiltonian, corresponding to a linear part of the CME Dynamics, a synchronously pumped parametric amplifier denoted here as PPL and Wave Guide adds a crucial nonlinear component to the CIA and Dynamics as well. In the basic CM algorithm, the pump power starts very low and has gradually increased at low pump powers. The amplitude of the easing spin pulses behaviors continuous, complex variables. Who Israel parts which can be positive or negative, play the role of play the role of soft or perhaps mean field spins once the pump, our crosses the threshold for parametric self oscillation. In the optical fiber ring, however, the attitudes of the easing spin pulses become effectively Qantas ized into binary values while the pump power is being ramped up. The F P J subsystem continuously applies its measurement based feedback. Implementation of the using Hamiltonian terms, the interplay of the linear rised using dynamics implemented by the F P G A and the threshold conversation dynamics provided by the sink pumped Parametric amplifier result in the final state of the optical optical pulse amplitude at the end of the pump ramp that could be read as a binary strain, giving a proposed solution of the easing ground state problem. This method of solving easing problem seems quite different from a conventional algorithm that runs entirely on a digital computer as a crucial aspect of the computation is performed physically by the analog, continuous, coherent, nonlinear dynamics of the optical degrees of freedom. In our efforts to analyze CIA and performance, we have therefore turned to the tools of dynamical systems theory, namely, a study of modifications, the evolution of critical points and apologies of hetero clinic orbits and basins of attraction. We conjecture that such analysis can provide fundamental insight into what makes certain optimization instances hard or easy for coherent using machines and hope that our approach can lead to both improvements of the course, the AM algorithm and a pre processing rubric for rapidly assessing the CME suitability of new instances. Okay, to provide a bit of intuition about how this all works, it may help to consider the threshold dynamics of just one or two optical parametric oscillators in the CME architecture just described. We can think of each of the pulse time slots circulating around the fiber ring, as are presenting an independent Opio. We can think of a single Opio degree of freedom as a single, resonant optical node that experiences linear dissipation, do toe out coupling loss and gain in a pump. Nonlinear crystal has shown in the diagram on the upper left of this slide as the pump power is increased from zero. As in the CME algorithm, the non linear game is initially to low toe overcome linear dissipation, and the Opio field remains in a near vacuum state at a critical threshold. Value gain. Equal participation in the Popeo undergoes a sort of lazing transition, and the study states of the OPIO above this threshold are essentially coherent states. There are actually two possible values of the Opio career in amplitude and any given above threshold pump power which are equal in magnitude but opposite in phase when the OPI across the special diet basically chooses one of the two possible phases randomly, resulting in the generation of a single bit of information. If we consider to uncoupled, Opio has shown in the upper right diagram pumped it exactly the same power at all times. Then, as the pump power has increased through threshold, each Opio will independently choose the phase and thus to random bits are generated for any number of uncoupled. Oppose the threshold power per opio is unchanged from the single Opio case. Now, however, consider a scenario in which the two appeals air, coupled to each other by a mutual injection of their out coupled fields has shown in the diagram on the lower right. One can imagine that depending on the sign of the coupling parameter Alfa, when one Opio is lazing, it will inject a perturbation into the other that may interfere either constructively or destructively, with the feel that it is trying to generate by its own lazing process. As a result, when came easily showed that for Alfa positive, there's an effective ferro magnetic coupling between the two Opio fields and their collective oscillation threshold is lowered from that of the independent Opio case. But on Lee for the two collective oscillation modes in which the two Opio phases are the same for Alfa Negative, the collective oscillation threshold is lowered on Lee for the configurations in which the Opio phases air opposite. So then, looking at how Alfa is related to the J. I. J matrix of the easing spin coupling Hamiltonian, it follows that we could use this simplistic to a p o. C. I am to solve the ground state problem of a fair magnetic or anti ferro magnetic ankles to easing model simply by increasing the pump power from zero and observing what phase relation occurs as the two appeals first start delays. Clearly, we can imagine generalizing this story toe larger, and however the story doesn't stay is clean and simple for all larger problem instances. And to find a more complicated example, we only need to go to n equals four for some choices of J J for n equals, for the story remains simple. Like the n equals two case. The figure on the upper left of this slide shows the energy of various critical points for a non frustrated and equals, for instance, in which the first bifurcated critical point that is the one that I forget to the lowest pump value a. Uh, this first bifurcated critical point flows as symptomatically into the lowest energy easing solution and the figure on the upper right. However, the first bifurcated critical point flows to a very good but sub optimal minimum at large pump power. The global minimum is actually given by a distinct critical critical point that first appears at a higher pump power and is not automatically connected to the origin. The basic C am algorithm is thus not able to find this global minimum. Such non ideal behaviors needs to become more confident. Larger end for the n equals 20 instance, showing the lower plots where the lower right plot is just a zoom into a region of the lower left lot. It can be seen that the global minimum corresponds to a critical point that first appears out of pump parameter, a around 0.16 at some distance from the idiomatic trajectory of the origin. That's curious to note that in both of these small and examples, however, the critical point corresponding to the global minimum appears relatively close to the idiomatic projector of the origin as compared to the most of the other local minima that appear. We're currently working to characterize the face portrait topology between the global minimum in the antibiotic trajectory of the origin, taking clues as to how the basic C am algorithm could be generalized to search for non idiomatic trajectories that jump to the global minimum during the pump ramp. Of course, n equals 20 is still too small to be of interest for practical optimization applications. But the advantage of beginning with the study of small instances is that we're able reliably to determine their global minima and to see how they relate to the 80 about trajectory of the origin in the basic C am algorithm. In the smaller and limit, we can also analyze fully quantum mechanical models of Syrian dynamics. But that's a topic for future talks. Um, existing large scale prototypes are pushing into the range of in equals 10 to the 4 10 to 5 to six. So our ultimate objective in theoretical analysis really has to be to try to say something about CIA and dynamics and regime of much larger in our initial approach to characterizing CIA and behavior in the large in regime relies on the use of random matrix theory, and this connects to prior research on spin classes, SK models and the tap equations etcetera. At present, we're focusing on statistical characterization of the CIA ingredient descent landscape, including the evolution of critical points in their Eigen value spectra. As the pump power is gradually increased. We're investigating, for example, whether there could be some way to exploit differences in the relative stability of the global minimum versus other local minima. We're also working to understand the deleterious or potentially beneficial effects of non ideologies, such as a symmetry in the implemented these and couplings. Looking one step ahead, we plan to move next in the direction of considering more realistic classes of problem instances such as quadratic, binary optimization with constraints. Eso In closing, I should acknowledge people who did the hard work on these things that I've shown eso. My group, including graduate students Ed winning, Daniel Wennberg, Tatsuya Nagamoto and Atsushi Yamamura, have been working in close collaboration with Syria Ganguly, Marty Fair and Amir Safarini Nini, all of us within the Department of Applied Physics at Stanford University. On also in collaboration with the Oshima Moto over at NTT 55 research labs, Onda should acknowledge funding support from the NSF by the Coherent Easing Machines Expedition in computing, also from NTT five research labs, Army Research Office and Exxon Mobil. Uh, that's it. Thanks very much. >>Mhm e >>t research and the Oshie for putting together this program and also the opportunity to speak here. My name is Al Gore ism or Andy and I'm from Caltech, and today I'm going to tell you about the work that we have been doing on networks off optical parametric oscillators and how we have been using them for icing machines and how we're pushing them toward Cornum photonics to acknowledge my team at Caltech, which is now eight graduate students and five researcher and postdocs as well as collaborators from all over the world, including entity research and also the funding from different places, including entity. So this talk is primarily about networks of resonate er's, and these networks are everywhere from nature. For instance, the brain, which is a network of oscillators all the way to optics and photonics and some of the biggest examples or metal materials, which is an array of small resonate er's. And we're recently the field of technological photonics, which is trying thio implement a lot of the technological behaviors of models in the condensed matter, physics in photonics and if you want to extend it even further, some of the implementations off quantum computing are technically networks of quantum oscillators. So we started thinking about these things in the context of icing machines, which is based on the icing problem, which is based on the icing model, which is the simple summation over the spins and spins can be their upward down and the couplings is given by the JJ. And the icing problem is, if you know J I J. What is the spin configuration that gives you the ground state? And this problem is shown to be an MP high problem. So it's computational e important because it's a representative of the MP problems on NPR. Problems are important because first, their heart and standard computers if you use a brute force algorithm and they're everywhere on the application side. That's why there is this demand for making a machine that can target these problems, and hopefully it can provide some meaningful computational benefit compared to the standard digital computers. So I've been building these icing machines based on this building block, which is a degenerate optical parametric. Oscillator on what it is is resonator with non linearity in it, and we pump these resonate er's and we generate the signal at half the frequency of the pump. One vote on a pump splits into two identical photons of signal, and they have some very interesting phase of frequency locking behaviors. And if you look at the phase locking behavior, you realize that you can actually have two possible phase states as the escalation result of these Opio which are off by pie, and that's one of the important characteristics of them. So I want to emphasize a little more on that and I have this mechanical analogy which are basically two simple pendulum. But there are parametric oscillators because I'm going to modulate the parameter of them in this video, which is the length of the string on by that modulation, which is that will make a pump. I'm gonna make a muscular. That'll make a signal which is half the frequency of the pump. And I have two of them to show you that they can acquire these face states so they're still facing frequency lock to the pump. But it can also lead in either the zero pie face states on. The idea is to use this binary phase to represent the binary icing spin. So each opio is going to represent spin, which can be either is your pie or up or down. And to implement the network of these resonate er's, we use the time off blood scheme, and the idea is that we put impulses in the cavity. These pulses air separated by the repetition period that you put in or t r. And you can think about these pulses in one resonator, xaz and temporarily separated synthetic resonate Er's if you want a couple of these resonator is to each other, and now you can introduce these delays, each of which is a multiple of TR. If you look at the shortest delay it couples resonator wanted to 2 to 3 and so on. If you look at the second delay, which is two times a rotation period, the couple's 123 and so on. And if you have and minus one delay lines, then you can have any potential couplings among these synthetic resonate er's. And if I can introduce these modulators in those delay lines so that I can strength, I can control the strength and the phase of these couplings at the right time. Then I can have a program will all toe all connected network in this time off like scheme, and the whole physical size of the system scales linearly with the number of pulses. So the idea of opium based icing machine is didn't having these o pos, each of them can be either zero pie and I can arbitrarily connect them to each other. And then I start with programming this machine to a given icing problem by just setting the couplings and setting the controllers in each of those delight lines. So now I have a network which represents an icing problem. Then the icing problem maps to finding the face state that satisfy maximum number of coupling constraints. And the way it happens is that the icing Hamiltonian maps to the linear loss of the network. And if I start adding gain by just putting pump into the network, then the OPI ohs are expected to oscillate in the lowest, lowest lost state. And, uh and we have been doing these in the past, uh, six or seven years and I'm just going to quickly show you the transition, especially what happened in the first implementation, which was using a free space optical system and then the guided wave implementation in 2016 and the measurement feedback idea which led to increasing the size and doing actual computation with these machines. So I just want to make this distinction here that, um, the first implementation was an all optical interaction. We also had an unequal 16 implementation. And then we transition to this measurement feedback idea, which I'll tell you quickly what it iss on. There's still a lot of ongoing work, especially on the entity side, to make larger machines using the measurement feedback. But I'm gonna mostly focused on the all optical networks and how we're using all optical networks to go beyond simulation of icing Hamiltonian both in the linear and non linear side and also how we're working on miniaturization of these Opio networks. So the first experiment, which was the four opium machine, it was a free space implementation and this is the actual picture off the machine and we implemented a small and it calls for Mexico problem on the machine. So one problem for one experiment and we ran the machine 1000 times, we looked at the state and we always saw it oscillate in one of these, um, ground states of the icing laboratoria. So then the measurement feedback idea was to replace those couplings and the controller with the simulator. So we basically simulated all those coherent interactions on on FB g. A. And we replicated the coherent pulse with respect to all those measurements. And then we injected it back into the cavity and on the near to you still remain. So it still is a non. They're dynamical system, but the linear side is all simulated. So there are lots of questions about if this system is preserving important information or not, or if it's gonna behave better. Computational wars. And that's still ah, lot of ongoing studies. But nevertheless, the reason that this implementation was very interesting is that you don't need the end minus one delight lines so you can just use one. Then you can implement a large machine, and then you can run several thousands of problems in the machine, and then you can compare the performance from the computational perspective Looks so I'm gonna split this idea of opium based icing machine into two parts. One is the linear part, which is if you take out the non linearity out of the resonator and just think about the connections. You can think about this as a simple matrix multiplication scheme. And that's basically what gives you the icing Hambletonian modeling. So the optical laws of this network corresponds to the icing Hamiltonian. And if I just want to show you the example of the n equals for experiment on all those face states and the history Graham that we saw, you can actually calculate the laws of each of those states because all those interferences in the beam splitters and the delay lines are going to give you a different losses. And then you will see that the ground states corresponds to the lowest laws of the actual optical network. If you add the non linearity, the simple way of thinking about what the non linearity does is that it provides to gain, and then you start bringing up the gain so that it hits the loss. Then you go through the game saturation or the threshold which is going to give you this phase bifurcation. So you go either to zero the pie face state. And the expectation is that Theis, the network oscillates in the lowest possible state, the lowest possible loss state. There are some challenges associated with this intensity Durban face transition, which I'm going to briefly talk about. I'm also going to tell you about other types of non aerodynamics that we're looking at on the non air side of these networks. So if you just think about the linear network, we're actually interested in looking at some technological behaviors in these networks. And the difference between looking at the technological behaviors and the icing uh, machine is that now, First of all, we're looking at the type of Hamilton Ian's that are a little different than the icing Hamilton. And one of the biggest difference is is that most of these technological Hamilton Ian's that require breaking the time reversal symmetry, meaning that you go from one spin to in the one side to another side and you get one phase. And if you go back where you get a different phase, and the other thing is that we're not just interested in finding the ground state, we're actually now interesting and looking at all sorts of states and looking at the dynamics and the behaviors of all these states in the network. So we started with the simplest implementation, of course, which is a one d chain of thes resonate, er's, which corresponds to a so called ssh model. In the technological work, we get the similar energy to los mapping and now we can actually look at the band structure on. This is an actual measurement that we get with this associate model and you see how it reasonably how How? Well, it actually follows the prediction and the theory. One of the interesting things about the time multiplexing implementation is that now you have the flexibility of changing the network as you are running the machine. And that's something unique about this time multiplex implementation so that we can actually look at the dynamics. And one example that we have looked at is we can actually go through the transition off going from top A logical to the to the standard nontrivial. I'm sorry to the trivial behavior of the network. You can then look at the edge states and you can also see the trivial and states and the technological at states actually showing up in this network. We have just recently implement on a two D, uh, network with Harper Hofstadter model and when you don't have the results here. But we're one of the other important characteristic of time multiplexing is that you can go to higher and higher dimensions and keeping that flexibility and dynamics, and we can also think about adding non linearity both in a classical and quantum regimes, which is going to give us a lot of exotic, no classical and quantum, non innate behaviors in these networks. Yeah, So I told you about the linear side. Mostly let me just switch gears and talk about the nonlinear side of the network. And the biggest thing that I talked about so far in the icing machine is this face transition that threshold. So the low threshold we have squeezed state in these. Oh, pios, if you increase the pump, we go through this intensity driven phase transition and then we got the face stays above threshold. And this is basically the mechanism off the computation in these O pos, which is through this phase transition below to above threshold. So one of the characteristics of this phase transition is that below threshold, you expect to see quantum states above threshold. You expect to see more classical states or coherent states, and that's basically corresponding to the intensity off the driving pump. So it's really hard to imagine that it can go above threshold. Or you can have this friends transition happen in the all in the quantum regime. And there are also some challenges associated with the intensity homogeneity off the network, which, for example, is if one opioid starts oscillating and then its intensity goes really high. Then it's going to ruin this collective decision making off the network because of the intensity driven face transition nature. So So the question is, can we look at other phase transitions? Can we utilize them for both computing? And also can we bring them to the quantum regime on? I'm going to specifically talk about the face transition in the spectral domain, which is the transition from the so called degenerate regime, which is what I mostly talked about to the non degenerate regime, which happens by just tuning the phase of the cavity. And what is interesting is that this phase transition corresponds to a distinct phase noise behavior. So in the degenerate regime, which we call it the order state, you're gonna have the phase being locked to the phase of the pump. As I talked about non degenerate regime. However, the phase is the phase is mostly dominated by the quantum diffusion. Off the off the phase, which is limited by the so called shallow towns limit, and you can see that transition from the general to non degenerate, which also has distinct symmetry differences. And this transition corresponds to a symmetry breaking in the non degenerate case. The signal can acquire any of those phases on the circle, so it has a you one symmetry. Okay, and if you go to the degenerate case, then that symmetry is broken and you only have zero pie face days I will look at. So now the question is can utilize this phase transition, which is a face driven phase transition, and can we use it for similar computational scheme? So that's one of the questions that were also thinking about. And it's not just this face transition is not just important for computing. It's also interesting from the sensing potentials and this face transition, you can easily bring it below threshold and just operated in the quantum regime. Either Gaussian or non Gaussian. If you make a network of Opio is now, we can see all sorts off more complicated and more interesting phase transitions in the spectral domain. One of them is the first order phase transition, which you get by just coupling to Opio, and that's a very abrupt face transition and compared to the to the single Opio phase transition. And if you do the couplings right, you can actually get a lot of non her mission dynamics and exceptional points, which are actually very interesting to explore both in the classical and quantum regime. And I should also mention that you can think about the cup links to be also nonlinear couplings. And that's another behavior that you can see, especially in the nonlinear in the non degenerate regime. So with that, I basically told you about these Opio networks, how we can think about the linear scheme and the linear behaviors and how we can think about the rich, nonlinear dynamics and non linear behaviors both in the classical and quantum regime. I want to switch gear and tell you a little bit about the miniaturization of these Opio networks. And of course, the motivation is if you look at the electron ICS and what we had 60 or 70 years ago with vacuum tube and how we transition from relatively small scale computers in the order of thousands of nonlinear elements to billions of non elements where we are now with the optics is probably very similar to 70 years ago, which is a table talk implementation. And the question is, how can we utilize nano photonics? I'm gonna just briefly show you the two directions on that which we're working on. One is based on lithium Diabate, and the other is based on even a smaller resonate er's could you? So the work on Nana Photonic lithium naive. It was started in collaboration with Harvard Marko Loncar, and also might affair at Stanford. And, uh, we could show that you can do the periodic polling in the phenomenon of it and get all sorts of very highly nonlinear processes happening in this net. Photonic periodically polls if, um Diabate. And now we're working on building. Opio was based on that kind of photonic the film Diabate. And these air some some examples of the devices that we have been building in the past few months, which I'm not gonna tell you more about. But the O. P. O. S. And the Opio Networks are in the works. And that's not the only way of making large networks. Um, but also I want to point out that The reason that these Nana photonic goblins are actually exciting is not just because you can make a large networks and it can make him compact in a in a small footprint. They also provide some opportunities in terms of the operation regime. On one of them is about making cat states and Opio, which is, can we have the quantum superposition of the zero pie states that I talked about and the Net a photonic within? I've It provides some opportunities to actually get closer to that regime because of the spatial temporal confinement that you can get in these wave guides. So we're doing some theory on that. We're confident that the type of non linearity two losses that it can get with these platforms are actually much higher than what you can get with other platform their existing platforms and to go even smaller. We have been asking the question off. What is the smallest possible Opio that you can make? Then you can think about really wavelength scale type, resonate er's and adding the chi to non linearity and see how and when you can get the Opio to operate. And recently, in collaboration with us see, we have been actually USC and Creole. We have demonstrated that you can use nano lasers and get some spin Hamilton and implementations on those networks. So if you can build the a P. O s, we know that there is a path for implementing Opio Networks on on such a nano scale. So we have looked at these calculations and we try to estimate the threshold of a pos. Let's say for me resonator and it turns out that it can actually be even lower than the type of bulk Pip Llano Pos that we have been building in the past 50 years or so. So we're working on the experiments and we're hoping that we can actually make even larger and larger scale Opio networks. So let me summarize the talk I told you about the opium networks and our work that has been going on on icing machines and the measurement feedback. And I told you about the ongoing work on the all optical implementations both on the linear side and also on the nonlinear behaviors. And I also told you a little bit about the efforts on miniaturization and going to the to the Nano scale. So with that, I would like Thio >>three from the University of Tokyo. Before I thought that would like to thank you showing all the stuff of entity for the invitation and the organization of this online meeting and also would like to say that it has been very exciting to see the growth of this new film lab. And I'm happy to share with you today of some of the recent works that have been done either by me or by character of Hong Kong. Honest Group indicates the title of my talk is a neuro more fic in silica simulator for the communities in machine. And here is the outline I would like to make the case that the simulation in digital Tektronix of the CME can be useful for the better understanding or improving its function principles by new job introducing some ideas from neural networks. This is what I will discuss in the first part and then it will show some proof of concept of the game and performance that can be obtained using dissimulation in the second part and the protection of the performance that can be achieved using a very large chaos simulator in the third part and finally talk about future plans. So first, let me start by comparing recently proposed izing machines using this table there is elected from recent natural tronics paper from the village Park hard people, and this comparison shows that there's always a trade off between energy efficiency, speed and scalability that depends on the physical implementation. So in red, here are the limitation of each of the servers hardware on, interestingly, the F p G, a based systems such as a producer, digital, another uh Toshiba beautification machine or a recently proposed restricted Bozeman machine, FPD A by a group in Berkeley. They offer a good compromise between speed and scalability. And this is why, despite the unique advantage that some of these older hardware have trust as the currency proposition in Fox, CBS or the energy efficiency off memory Sisters uh P. J. O are still an attractive platform for building large organizing machines in the near future. The reason for the good performance of Refugee A is not so much that they operate at the high frequency. No, there are particular in use, efficient, but rather that the physical wiring off its elements can be reconfigured in a way that limits the funding human bottleneck, larger, funny and phenols and the long propagation video information within the system. In this respect, the LPGA is They are interesting from the perspective off the physics off complex systems, but then the physics of the actions on the photos. So to put the performance of these various hardware and perspective, we can look at the competition of bringing the brain the brain complete, using billions of neurons using only 20 watts of power and operates. It's a very theoretically slow, if we can see and so this impressive characteristic, they motivate us to try to investigate. What kind of new inspired principles be useful for designing better izing machines? The idea of this research project in the future collaboration it's to temporary alleviates the limitations that are intrinsic to the realization of an optical cortex in machine shown in the top panel here. By designing a large care simulator in silicone in the bottom here that can be used for digesting the better organization principles of the CIA and this talk, I will talk about three neuro inspired principles that are the symmetry of connections, neural dynamics orphan chaotic because of symmetry, is interconnectivity the infrastructure? No. Next talks are not composed of the reputation of always the same types of non environments of the neurons, but there is a local structure that is repeated. So here's the schematic of the micro column in the cortex. And lastly, the Iraqi co organization of connectivity connectivity is organizing a tree structure in the brain. So here you see a representation of the Iraqi and organization of the monkey cerebral cortex. So how can these principles we used to improve the performance of the icing machines? And it's in sequence stimulation. So, first about the two of principles of the estimate Trian Rico structure. We know that the classical approximation of the car testing machine, which is the ground toe, the rate based on your networks. So in the case of the icing machines, uh, the okay, Scott approximation can be obtained using the trump active in your position, for example, so the times of both of the system they are, they can be described by the following ordinary differential equations on in which, in case of see, I am the X, I represent the in phase component of one GOP Oh, Theo f represents the monitor optical parts, the district optical Parametric amplification and some of the good I JoJo extra represent the coupling, which is done in the case of the measure of feedback coupling cm using oh, more than detection and refugee A and then injection off the cooking time and eso this dynamics in both cases of CNN in your networks, they can be written as the grand set of a potential function V, and this written here, and this potential functionally includes the rising Maccagnan. So this is why it's natural to use this type of, uh, dynamics to solve the icing problem in which the Omega I J or the eyes in coping and the H is the extension of the icing and attorney in India and expect so. Not that this potential function can only be defined if the Omega I j. R. A. Symmetric. So the well known problem of this approach is that this potential function V that we obtain is very non convicts at low temperature, and also one strategy is to gradually deformed this landscape, using so many in process. But there is no theorem. Unfortunately, that granted conventions to the global minimum of There's even Tony and using this approach. And so this is why we propose, uh, to introduce a macro structures of the system where one analog spin or one D O. P. O is replaced by a pair off one another spin and one error, according viable. And the addition of this chemical structure introduces a symmetry in the system, which in terms induces chaotic dynamics, a chaotic search rather than a learning process for searching for the ground state of the icing. Every 20 within this massacre structure the role of the er variable eyes to control the amplitude off the analog spins toe force. The amplitude of the expense toe become equal to certain target amplitude a uh and, uh, and this is done by modulating the strength off the icing complaints or see the the error variable E I multiply the icing complaint here in the dynamics off air d o p. O. On then the dynamics. The whole dynamics described by this coupled equations because the e I do not necessarily take away the same value for the different. I thesis introduces a symmetry in the system, which in turn creates security dynamics, which I'm sure here for solving certain current size off, um, escape problem, Uh, in which the X I are shown here and the i r from here and the value of the icing energy showing the bottom plots. You see this Celtics search that visit various local minima of the as Newtonian and eventually finds the global minimum? Um, it can be shown that this modulation off the target opportunity can be used to destabilize all the local minima off the icing evertonians so that we're gonna do not get stuck in any of them. On more over the other types of attractors I can eventually appear, such as limits I contractors, Okot contractors. They can also be destabilized using the motivation of the target and Batuta. And so we have proposed in the past two different moderation of the target amateur. The first one is a modulation that ensure the uh 100 reproduction rate of the system to become positive on this forbids the creation off any nontrivial tractors. And but in this work, I will talk about another moderation or arrested moderation which is given here. That works, uh, as well as this first uh, moderation, but is easy to be implemented on refugee. So this couple of the question that represent becoming the stimulation of the cortex in machine with some error correction they can be implemented especially efficiently on an F B. G. And here I show the time that it takes to simulate three system and also in red. You see, at the time that it takes to simulate the X I term the EI term, the dot product and the rising Hamiltonian for a system with 500 spins and Iraq Spain's equivalent to 500 g. O. P. S. So >>in >>f b d a. The nonlinear dynamics which, according to the digital optical Parametric amplification that the Opa off the CME can be computed in only 13 clock cycles at 300 yards. So which corresponds to about 0.1 microseconds. And this is Toby, uh, compared to what can be achieved in the measurements back O C. M. In which, if we want to get 500 timer chip Xia Pios with the one she got repetition rate through the obstacle nine narrative. Uh, then way would require 0.5 microseconds toe do this so the submission in F B J can be at least as fast as ah one g repression. Uh, replicate pulsed laser CIA Um, then the DOT product that appears in this differential equation can be completed in 43 clock cycles. That's to say, one microseconds at 15 years. So I pieced for pouring sizes that are larger than 500 speeds. The dot product becomes clearly the bottleneck, and this can be seen by looking at the the skating off the time the numbers of clock cycles a text to compute either the non in your optical parts or the dog products, respect to the problem size. And And if we had infinite amount of resources and PGA to simulate the dynamics, then the non illogical post can could be done in the old one. On the mattress Vector product could be done in the low carrot off, located off scales as a look at it off and and while the guide off end. Because computing the dot product involves assuming all the terms in the product, which is done by a nephew, GE by another tree, which heights scarce logarithmic any with the size of the system. But This is in the case if we had an infinite amount of resources on the LPGA food, but for dealing for larger problems off more than 100 spins. Usually we need to decompose the metrics into ah, smaller blocks with the block side that are not you here. And then the scaling becomes funny, non inner parts linear in the end, over you and for the products in the end of EU square eso typically for low NF pdf cheap PGA you the block size off this matrix is typically about 100. So clearly way want to make you as large as possible in order to maintain this scanning in a log event for the numbers of clock cycles needed to compute the product rather than this and square that occurs if we decompose the metrics into smaller blocks. But the difficulty in, uh, having this larger blocks eyes that having another tree very large Haider tree introduces a large finding and finance and long distance start a path within the refugee. So the solution to get higher performance for a simulator of the contest in machine eyes to get rid of this bottleneck for the dot product by increasing the size of this at the tree. And this can be done by organizing your critique the electrical components within the LPGA in order which is shown here in this, uh, right panel here in order to minimize the finding finance of the system and to minimize the long distance that a path in the in the fpt So I'm not going to the details of how this is implemented LPGA. But just to give you a idea off why the Iraqi Yahiko organization off the system becomes the extremely important toe get good performance for similar organizing machine. So instead of instead of getting into the details of the mpg implementation, I would like to give some few benchmark results off this simulator, uh, off the that that was used as a proof of concept for this idea which is can be found in this archive paper here and here. I should results for solving escape problems. Free connected person, randomly person minus one spring last problems and we sure, as we use as a metric the numbers of the mattress Victor products since it's the bottleneck of the computation, uh, to get the optimal solution of this escape problem with the Nina successful BT against the problem size here and and in red here, this propose FDJ implementation and in ah blue is the numbers of retrospective product that are necessary for the C. I am without error correction to solve this escape programs and in green here for noisy means in an evening which is, uh, behavior with similar to the Cartesian mission. Uh, and so clearly you see that the scaring off the numbers of matrix vector product necessary to solve this problem scales with a better exponents than this other approaches. So So So that's interesting feature of the system and next we can see what is the real time to solution to solve this SK instances eso in the last six years, the time institution in seconds to find a grand state of risk. Instances remain answers probability for different state of the art hardware. So in red is the F B g. A presentation proposing this paper and then the other curve represent Ah, brick a local search in in orange and silver lining in purple, for example. And so you see that the scaring off this purpose simulator is is rather good, and that for larger plant sizes we can get orders of magnitude faster than the state of the art approaches. Moreover, the relatively good scanning off the time to search in respect to problem size uh, they indicate that the FPD implementation would be faster than risk. Other recently proposed izing machine, such as the hope you know, natural complimented on memories distance that is very fast for small problem size in blue here, which is very fast for small problem size. But which scanning is not good on the same thing for the restricted Bosman machine. Implementing a PGA proposed by some group in Broken Recently Again, which is very fast for small parliament sizes but which canning is bad so that a dis worse than the proposed approach so that we can expect that for programs size is larger than 1000 spins. The proposed, of course, would be the faster one. Let me jump toe this other slide and another confirmation that the scheme scales well that you can find the maximum cut values off benchmark sets. The G sets better candidates that have been previously found by any other algorithms, so they are the best known could values to best of our knowledge. And, um or so which is shown in this paper table here in particular, the instances, uh, 14 and 15 of this G set can be We can find better converse than previously known, and we can find this can vary is 100 times faster than the state of the art algorithm and CP to do this which is a very common Kasich. It s not that getting this a good result on the G sets, they do not require ah, particular hard tuning of the parameters. So the tuning issuing here is very simple. It it just depends on the degree off connectivity within each graph. And so this good results on the set indicate that the proposed approach would be a good not only at solving escape problems in this problems, but all the types off graph sizing problems on Mexican province in communities. So given that the performance off the design depends on the height of this other tree, we can try to maximize the height of this other tree on a large F p g a onda and carefully routing the components within the P G A and and we can draw some projections of what type of performance we can achieve in the near future based on the, uh, implementation that we are currently working. So here you see projection for the time to solution way, then next property for solving this escape programs respect to the prime assize. And here, compared to different with such publicizing machines, particularly the digital. And, you know, 42 is shown in the green here, the green line without that's and, uh and we should two different, uh, hypothesis for this productions either that the time to solution scales as exponential off n or that the time of social skills as expression of square root off. So it seems, according to the data, that time solution scares more as an expression of square root of and also we can be sure on this and this production show that we probably can solve prime escape problem of science 2000 spins, uh, to find the rial ground state of this problem with 99 success ability in about 10 seconds, which is much faster than all the other proposed approaches. So one of the future plans for this current is in machine simulator. So the first thing is that we would like to make dissimulation closer to the rial, uh, GOP oh, optical system in particular for a first step to get closer to the system of a measurement back. See, I am. And to do this what is, uh, simulate Herbal on the p a is this quantum, uh, condoms Goshen model that is proposed described in this paper and proposed by people in the in the Entity group. And so the idea of this model is that instead of having the very simple or these and have shown previously, it includes paired all these that take into account on me the mean off the awesome leverage off the, uh, European face component, but also their violence s so that we can take into account more quantum effects off the g o p. O, such as the squeezing. And then we plan toe, make the simulator open access for the members to run their instances on the system. There will be a first version in September that will be just based on the simple common line access for the simulator and in which will have just a classic or approximation of the system. We don't know Sturm, binary weights and museum in term, but then will propose a second version that would extend the current arising machine to Iraq off F p g. A, in which we will add the more refined models truncated, ignoring the bottom Goshen model they just talked about on the support in which he valued waits for the rising problems and support the cement. So we will announce later when this is available and and far right is working >>hard comes from Universal down today in physics department, and I'd like to thank the organizers for their kind invitation to participate in this very interesting and promising workshop. Also like to say that I look forward to collaborations with with a file lab and Yoshi and collaborators on the topics of this world. So today I'll briefly talk about our attempt to understand the fundamental limits off another continues time computing, at least from the point off you off bullion satisfy ability, problem solving, using ordinary differential equations. But I think the issues that we raise, um, during this occasion actually apply to other other approaches on a log approaches as well and into other problems as well. I think everyone here knows what Dorien satisfy ability. Problems are, um, you have boolean variables. You have em clauses. Each of disjunction of collaterals literally is a variable, or it's, uh, negation. And the goal is to find an assignment to the variable, such that order clauses are true. This is a decision type problem from the MP class, which means you can checking polynomial time for satisfy ability off any assignment. And the three set is empty, complete with K three a larger, which means an efficient trees. That's over, uh, implies an efficient source for all the problems in the empty class, because all the problems in the empty class can be reduced in Polian on real time to reset. As a matter of fact, you can reduce the NP complete problems into each other. You can go from three set to set backing or two maximum dependent set, which is a set packing in graph theoretic notions or terms toe the icing graphs. A problem decision version. This is useful, and you're comparing different approaches, working on different kinds of problems when not all the closest can be satisfied. You're looking at the accusation version offset, uh called Max Set. And the goal here is to find assignment that satisfies the maximum number of clauses. And this is from the NPR class. In terms of applications. If we had inefficient sets over or np complete problems over, it was literally, positively influenced. Thousands off problems and applications in industry and and science. I'm not going to read this, but this this, of course, gives a strong motivation toe work on this kind of problems. Now our approach to set solving involves embedding the problem in a continuous space, and you use all the east to do that. So instead of working zeros and ones, we work with minus one across once, and we allow the corresponding variables toe change continuously between the two bounds. We formulate the problem with the help of a close metrics. If if a if a close, uh, does not contain a variable or its negation. The corresponding matrix element is zero. If it contains the variable in positive, for which one contains the variable in a gated for Mitt's negative one, and then we use this to formulate this products caused quote, close violation functions one for every clause, Uh, which really, continuously between zero and one. And they're zero if and only if the clause itself is true. Uh, then we form the define in order to define a dynamic such dynamics in this and dimensional hyper cube where the search happens and if they exist, solutions. They're sitting in some of the corners of this hyper cube. So we define this, uh, energy potential or landscape function shown here in a way that this is zero if and only if all the clauses all the kmc zero or the clauses off satisfied keeping these auxiliary variables a EMS always positive. And therefore, what you do here is a dynamics that is a essentially ingredient descend on this potential energy landscape. If you were to keep all the M's constant that it would get stuck in some local minimum. However, what we do here is we couple it with the dynamics we cooperated the clothes violation functions as shown here. And if he didn't have this am here just just the chaos. For example, you have essentially what case you have positive feedback. You have increasing variable. Uh, but in that case, you still get stuck would still behave will still find. So she is better than the constant version but still would get stuck only when you put here this a m which makes the dynamics in in this variable exponential like uh, only then it keeps searching until he finds a solution on deer is a reason for that. I'm not going toe talk about here, but essentially boils down toe performing a Grady and descend on a globally time barren landscape. And this is what works. Now I'm gonna talk about good or bad and maybe the ugly. Uh, this is, uh, this is What's good is that it's a hyperbolic dynamical system, which means that if you take any domain in the search space that doesn't have a solution in it or any socially than the number of trajectories in it decays exponentially quickly. And the decay rate is a characteristic in variant characteristic off the dynamics itself. Dynamical systems called the escape right the inverse off that is the time scale in which you find solutions by this by this dynamical system, and you can see here some song trajectories that are Kelty because it's it's no linear, but it's transient, chaotic. Give their sources, of course, because eventually knowledge to the solution. Now, in terms of performance here, what you show for a bunch off, um, constraint densities defined by M overran the ratio between closes toe variables for random, said Problems is random. Chris had problems, and they as its function off n And we look at money toward the wartime, the wall clock time and it behaves quite value behaves Azat party nominally until you actually he to reach the set on set transition where the hardest problems are found. But what's more interesting is if you monitor the continuous time t the performance in terms off the A narrow, continuous Time t because that seems to be a polynomial. And the way we show that is, we consider, uh, random case that random three set for a fixed constraint density Onda. We hear what you show here. Is that the right of the trash hold that it's really hard and, uh, the money through the fraction of problems that we have not been able to solve it. We select thousands of problems at that constraint ratio and resolve them without algorithm, and we monitor the fractional problems that have not yet been solved by continuous 90. And this, as you see these decays exponentially different. Educate rates for different system sizes, and in this spot shows that is dedicated behaves polynomial, or actually as a power law. So if you combine these two, you find that the time needed to solve all problems except maybe appear traction off them scales foreign or merely with the problem size. So you have paranormal, continuous time complexity. And this is also true for other types of very hard constraints and sexual problems such as exact cover, because you can always transform them into three set as we discussed before, Ramsey coloring and and on these problems, even algorithms like survey propagation will will fail. But this doesn't mean that P equals NP because what you have first of all, if you were toe implement these equations in a device whose behavior is described by these, uh, the keys. Then, of course, T the continue style variable becomes a physical work off. Time on that will be polynomial is scaling, but you have another other variables. Oxidative variables, which structured in an exponential manner. So if they represent currents or voltages in your realization and it would be an exponential cost Al Qaeda. But this is some kind of trade between time and energy, while I know how toe generate energy or I don't know how to generate time. But I know how to generate energy so it could use for it. But there's other issues as well, especially if you're trying toe do this son and digital machine but also happens. Problems happen appear. Other problems appear on in physical devices as well as we discuss later. So if you implement this in GPU, you can. Then you can get in order off to magnitude. Speed up. And you can also modify this to solve Max sad problems. Uh, quite efficiently. You are competitive with the best heuristic solvers. This is a weather problems. In 2016 Max set competition eso so this this is this is definitely this seems like a good approach, but there's off course interesting limitations, I would say interesting, because it kind of makes you think about what it means and how you can exploit this thes observations in understanding better on a low continues time complexity. If you monitored the discrete number the number of discrete steps. Don't buy the room, Dakota integrator. When you solve this on a digital machine, you're using some kind of integrator. Um and you're using the same approach. But now you measure the number off problems you haven't sold by given number of this kid, uh, steps taken by the integrator. You find out you have exponential, discrete time, complexity and, of course, thistles. A problem. And if you look closely, what happens even though the analog mathematical trajectory, that's the record here. If you monitor what happens in discrete time, uh, the integrator frustrates very little. So this is like, you know, third or for the disposition, but fluctuates like crazy. So it really is like the intervention frees us out. And this is because of the phenomenon of stiffness that are I'll talk a little bit a more about little bit layer eso. >>You know, it might look >>like an integration issue on digital machines that you could improve and could definitely improve. But actually issues bigger than that. It's It's deeper than that, because on a digital machine there is no time energy conversion. So the outside variables are efficiently representing a digital machine. So there's no exponential fluctuating current of wattage in your computer when you do this. Eso If it is not equal NP then the exponential time, complexity or exponential costs complexity has to hit you somewhere. And this is how um, but, you know, one would be tempted to think maybe this wouldn't be an issue in a analog device, and to some extent is true on our devices can be ordered to maintain faster, but they also suffer from their own problems because he not gonna be affect. That classes soldiers as well. So, indeed, if you look at other systems like Mirandizing machine measurement feedback, probably talk on the grass or selected networks. They're all hinge on some kind off our ability to control your variables in arbitrary, high precision and a certain networks you want toe read out across frequencies in case off CM's. You required identical and program because which is hard to keep, and they kind of fluctuate away from one another, shift away from one another. And if you control that, of course that you can control the performance. So actually one can ask if whether or not this is a universal bottleneck and it seems so aside, I will argue next. Um, we can recall a fundamental result by by showing harder in reaction Target from 1978. Who says that it's a purely computer science proof that if you are able toe, compute the addition multiplication division off riel variables with infinite precision, then you could solve any complete problems in polynomial time. It doesn't actually proposals all where he just chose mathematically that this would be the case. Now, of course, in Real warned, you have also precision. So the next question is, how does that affect the competition about problems? This is what you're after. Lots of precision means information also, or entropy production. Eso what you're really looking at the relationship between hardness and cost of computing off a problem. Uh, and according to Sean Hagar, there's this left branch which in principle could be polynomial time. But the question whether or not this is achievable that is not achievable, but something more cheerful. That's on the right hand side. There's always going to be some information loss, so mental degeneration that could keep you away from possibly from point normal time. So this is what we like to understand, and this information laws the source off. This is not just always I will argue, uh, in any physical system, but it's also off algorithm nature, so that is a questionable area or approach. But China gets results. Security theoretical. No, actual solar is proposed. So we can ask, you know, just theoretically get out off. Curiosity would in principle be such soldiers because it is not proposing a soldier with such properties. In principle, if if you want to look mathematically precisely what the solar does would have the right properties on, I argue. Yes, I don't have a mathematical proof, but I have some arguments that that would be the case. And this is the case for actually our city there solver that if you could calculate its trajectory in a loss this way, then it would be, uh, would solve epic complete problems in polynomial continuous time. Now, as a matter of fact, this a bit more difficult question, because time in all these can be re scared however you want. So what? Burns says that you actually have to measure the length of the trajectory, which is a new variant off the dynamical system or property dynamical system, not off its parameters ization. And we did that. So Suba Corral, my student did that first, improving on the stiffness off the problem off the integrations, using implicit solvers and some smart tricks such that you actually are closer to the actual trajectory and using the same approach. You know what fraction off problems you can solve? We did not give the length of the trajectory. You find that it is putting on nearly scaling the problem sites we have putting on your skin complexity. That means that our solar is both Polly length and, as it is, defined it also poorly time analog solver. But if you look at as a discreet algorithm, if you measure the discrete steps on a digital machine, it is an exponential solver. And the reason is because off all these stiffness, every integrator has tow truck it digitizing truncate the equations, and what it has to do is to keep the integration between the so called stability region for for that scheme, and you have to keep this product within a grimace of Jacoby in and the step size read in this region. If you use explicit methods. You want to stay within this region? Uh, but what happens that some off the Eigen values grow fast for Steve problems, and then you're you're forced to reduce that t so the product stays in this bonded domain, which means that now you have to you're forced to take smaller and smaller times, So you're you're freezing out the integration and what I will show you. That's the case. Now you can move to increase its soldiers, which is which is a tree. In this case, you have to make domain is actually on the outside. But what happens in this case is some of the Eigen values of the Jacobean, also, for six systems, start to move to zero. As they're moving to zero, they're going to enter this instability region, so your soul is going to try to keep it out, so it's going to increase the data T. But if you increase that to increase the truncation hours, so you get randomized, uh, in the large search space, so it's it's really not, uh, not going to work out. Now, one can sort off introduce a theory or language to discuss computational and are computational complexity, using the language from dynamical systems theory. But basically I I don't have time to go into this, but you have for heart problems. Security object the chaotic satellite Ouch! In the middle of the search space somewhere, and that dictates how the dynamics happens and variant properties off the dynamics. Of course, off that saddle is what the targets performance and many things, so a new, important measure that we find that it's also helpful in describing thesis. Another complexity is the so called called Makarov, or metric entropy and basically what this does in an intuitive A eyes, uh, to describe the rate at which the uncertainty containing the insignificant digits off a trajectory in the back, the flow towards the significant ones as you lose information because off arrows being, uh grown or are developed in tow. Larger errors in an exponential at an exponential rate because you have positively up north spawning. But this is an in variant property. It's the property of the set of all. This is not how you compute them, and it's really the interesting create off accuracy philosopher dynamical system. A zay said that you have in such a high dimensional that I'm consistent were positive and negatively upon of exponents. Aziz Many The total is the dimension of space and user dimension, the number off unstable manifold dimensions and as Saddam was stable, manifold direction. And there's an interesting and I think, important passion, equality, equality called the passion, equality that connect the information theoretic aspect the rate off information loss with the geometric rate of which trajectory separate minus kappa, which is the escape rate that I already talked about. Now one can actually prove a simple theorems like back off the envelope calculation. The idea here is that you know the rate at which the largest rated, which closely started trajectory separate from one another. So now you can say that, uh, that is fine, as long as my trajectory finds the solution before the projective separate too quickly. In that case, I can have the hope that if I start from some region off the face base, several close early started trajectories, they kind of go into the same solution orphaned and and that's that's That's this upper bound of this limit, and it is really showing that it has to be. It's an exponentially small number. What? It depends on the end dependence off the exponents right here, which combines information loss rate and the social time performance. So these, if this exponents here or that has a large independence or river linear independence, then you then you really have to start, uh, trajectories exponentially closer to one another in orderto end up in the same order. So this is sort off like the direction that you're going in tow, and this formulation is applicable toe all dynamical systems, uh, deterministic dynamical systems. And I think we can We can expand this further because, uh, there is, ah, way off getting the expression for the escaped rate in terms off n the number of variables from cycle expansions that I don't have time to talk about. What? It's kind of like a program that you can try toe pursuit, and this is it. So the conclusions I think of self explanatory I think there is a lot of future in in, uh, in an allo. Continue start computing. Um, they can be efficient by orders of magnitude and digital ones in solving empty heart problems because, first of all, many of the systems you like the phone line and bottleneck. There's parallelism involved, and and you can also have a large spectrum or continues time, time dynamical algorithms than discrete ones. And you know. But we also have to be mindful off. What are the possibility of what are the limits? And 11 open question is very important. Open question is, you know, what are these limits? Is there some kind off no go theory? And that tells you that you can never perform better than this limit or that limit? And I think that's that's the exciting part toe to derive thes thes this levian 10.

from the cube studios in Palo Alto in Boston it's the cube covering empowering the autonomous enterprise brought to you by Oracle consulting welcome back everybody to this special digital event coverage that the cube is looking into the rebirth of Oracle consulting Janet George is here she's group vp autonomous for advanced analytics with machine learning and artificial intelligence at oracle and she's joined by grant gibson is a group vp of growth and strategy at oracle folks welcome to the cube thanks so much for coming on thank you thank you great I want to start with you because you get strategy in your title like just start big picture what is the strategy with Oracle specifically as it relates to autonomous and also consulting sure so I think you know Oracle has a deep legacy of strengthened data and over the company's successful history it's evolved what that is from steps along the way if you look at the modern enterprise of Oracle client I think there's no denying that we've entered the age of AI that everyone knows that artificial intelligence and machine learning are a key to their success in the business marketplace going forward and while generally it's acknowledge that it's a transformative technology and people know that they need to take advantage of it it's the how that's really tricky and that most enterprises in order to really get an enterprise level ROI on an AI investment need to engage in projects of significant scope and going from realizing there's an opportunity to realize and there's a threat to mobilizing yourself to capitalize on it is a is a daunting task for an enemy certainly one that's you know anybody that's got any sort of legacy of success has built-in processes that's built in systems has built in skillsets and making that leap to be an autonomous enterprise is is challenging for companies to wrap their heads around so as part of the rebirth of Oracle consulting we've developed a practice around how to both manage the the technology needs for that transformation as well as the human needs as well as the data science needs to it so rather there's about five or six things that I want to followup with you there so there's gonna be good conversations Janet so ever since I've been in the industry we're talking about AI in sort of start stop start stop we had the AI winter and now it seems to be here it's almost feel like that the the technology never lived up to its promise you didn't have the horsepower a compute power you know enough data maybe so we're here today feels like we are entering a new era why is that and and how will the technology perform this time so for AI to perform it's very reliant on the data we entered the age of AI without having the right data for AI so you can imagine that we we just launched into AI without our data being ready to be training sex for AI so we started with bi data or we started the data that was already historically transformed formatted had logical structures physical structures this data was sort of trapped in many different tools and then suddenly AI comes along and we say take this data our historical data we haven't tested to see if this has labels in it this has learning capability in it we just thrust the data to AI and that's why we saw the initial wave of AI sort of failing because it was not ready to fall AI ready for the generation of AI and part of I think the leap that clients are finding success with now is getting the Apple data types and you're moving from the zeros and ones of structured data to image language written language spoken language you're capturing different data sets in ways that prior tools never could and so the classifications that come out of it the insights that come out of it the business process transformation comes out of it is different than what we would have understood under the structured data format so I think it's that combination of really being able to push massive amounts of data through a cloud product to be able to process it at scale that is what I think is the combination that takes it to the next plateau for sure the language that we use today I feel like is going to change and you just started to touch on some of them you know sensing you know they're our senses and you know the visualization and the the the the auditory so it's it's sort of this new experience that customers are saying a lot of this machine intelligence behind them I call it the autonomous enterprise right the journey to be the autonomous enterprise and when you're on this journey to be the autonomous enterprise you need really the platform that can help you be cloud is that platform which can help you get to the autonomous journey but the autonomous journey does not end with the cloud right or doesn't end with the dead lake these are just infrastructures that are basic necessary necessities for being on that on that autonomous journey but at the end it's about how do you train and scale at a very large scale training that needs to happen on this platform for AI to be successful and if you are an autonomous enterprise then you have really figured out how to tap into AI and machine learning in a way that nobody else has to derive business value if you will so you've got the platform you've got the data and now you're actually tapping into the autonomous components AI and machine learning to derive business intelligence and business value so I want to get into a little bit of Oracle's role but to do that I want to talk a little bit more about the industry so if you think about the way this the industry seems to be restructuring around data there historically Industries had their own stack or value chain and if you were in the finance industry you were there for life you know so when you think about banking for example highly regulated industry think about our geek culture these are highly regulated industries they're come it was very difficult to disrupt these industries but now you look at an Amazon right and what does an Amazon or any other tech giant like Apple have they have incredible amounts of data they understand how people use or how they want to do banking and so they've cut off the tap of cash or Amazon pay and these things are starting to eat into the market right so you would have never thought an Amazon could be a competition to your banking industry just because of regulations but they are not hindered by the regulations because they're starting at a different level and so they become an instant threat and an instant destructor to these highly regulated industries that's what data does right then you use data as you DNA for your business and you are sort of born in data or you figured out how to be autonomous if you will capture value from that data in a very significant manner then you can get into industries that are not traditionally your own industry it can be like the food industry it can be the cloud industry the book industry you know different industries so you know that that's what I see happening with the tech giants so great this is a really interesting point that Gina is making that you mentioned you started off with like a couple of industries that are highly regulated harder to disrupt you know music got disrupted publishing got disrupted but you've got these regulated businesses you know defense automotive actually hasn't been truly disrupted yet so I'm Tesla maybes a harbinger and so you've got this spectrum of disruption but is anybody safe from disruption okay I don't think anyone's ever safe from it it's it's changed in evolution right that you whether it's you know swapping horseshoes for cars or TV for movies or Netflix or any sort of evolution of a business you I wouldn't coast on any of them and I think to earlier question around the value that we can help bring to Oracle customers is that you know we have a rich stack of applications and I find that the space between the applications the data that that spans more than one of them is a ripe playground for innovations that where the data already exists inside a company but it's trapped from both a technology and a business perspective and that's where I think really any company can take advantage of knowing its data better and changing itself to take advantage of what's already there yet powerful bit people always throw the bromide out the data is the new oil and we've said no data is far more valuable because you can use it in a lot of different places or you can use once and it's has to follow laws of scarcity data if you can unlock it and so a lot of the incumbents they have built a business around whatever a factory or you know process and people a lot of the the trillion-dollar start in us that they're become trillionaires you know I'm talking about data is at the core their data company so so it seems like a big challenge for you you're incumbent customers clients is to put data hit the core be able to break down those silos how do they do that grading down silos is really super critical for any business it was okay to operate in a silo for example you would think that oh you know I could just be payroll in expense reports and it wouldn't man matter if I get into vendor performance management or purchasing that can operate as a silo but anymore we are finding that there are tremendous insights between vendor performance management I expensive all these things are all connected so you can't afford to have your data set in silos so grading down that silo actually gives the business very good performance right insights that they didn't have before so that's one way to go but but another phenomena happens when you start to great down the silos you start to recognize what data you don't have to take your business to the next level right that awareness will not happen when you're working with existing data so that awareness comes into form when you great the silos and you start to figure out you need to go after different set of data to get you to new product creation what would that look like new test insights or new capex avoidance then that data is just you have to go through the eye tration to be able to figure that out which takes is what you're saying happy so this notion of the autonomous under president help me here because I get kind of autonomous and automation coming into IT IT ops I'm interested in how you see customers taking that beyond the technology organization into the enterprise I think when AI is a technology problem the company is it at a loss ai has to be a business problem ai has to inform the business strategy ai has two main companies the successful companies that have done so 90 percent of our investments are going towards data we know that and and most of it going towards AI data out there about this right and so we looked at what are these ninety cup ninety percent of the company's investments where are these going and who is doing this right and who's not doing this right one of the things we are seeing as results is that the companies that are doing it right have brought data into their business strategy they've changed their business model right so it's not like making a better taxi but coming up with uber right so it's not like saying okay I'm going to have all these I'm going to be the drug manufacturing company I'm going to put drugs out there in the market versus I'm going to do connected health right and so how does data serve the business model of being connected health rather than being a drug company selling drugs to my customers right it's a completely different way of looking at it and so now I is informing drug discovery AI is not helping you just put more drugs to the market rather it's helping you come up with new drugs that will help the process of connected game there's a lot of discussion in the press about you know the ethics of AI and how far should we take AI and how far can we take it from a technology standpoint long roadmap there but how far should we take it do you feel as though public policy will take care of that a lot of that narrative is just kind of journalists looking for you know the negative story well that's sort itself out how much time do you spend with your customers talking about that we in Oracle we're building our data science platform with an explicit feature called explain ability off the model on how the model came up with the features what features it picked we can rearrange the features that the model picked so I think explain ability is very important for ordinary people to trust AI because we can't trust AI even even data scientists contrast AI right to a large extent so for us to get to that level where we can really trust what AI is picking in terms of a model we need to have explained ability and I think a lot of the companies right now are starting to make that as part of their platform well we're definitely entering a new era the the age of AI of the autonomous enterprise folks thanks very much for a great segment really appreciate it yeah our pleasure thank you for having us thank you alright and thank you and keep it right there we're right back with our next guest for this short break you're watching the cubes coverage of the rebirth of Oracle consulting right back you [Music]

I'm going alive I'm live right now let's send you this link and see if you can get on here so this is private see if I can break this out this is [Music] [Music] [Music] [Music] hello they're coming you live from Chuck alley studio here in Mountain View California and I'm on YouTube live I hope I'm not securing anything outta been out there for two minutes now let's be able to do a live private stream and be able to have that account that link to people - yeah okay yes you see me voice what's up what's up what's up so this is a private link I don't know if you can hear me that's a private link and if you give the link to whoever you want to see it oh you can't hear me hmm one two one two one two three four stop that

from the cube studios in Palo Alto in Boston connecting with thought leaders all around the world this is a cube conversation I am stupid a man and this is a special cube conversation we've been talking a lot of course for many years about the ascent of cloud and today in 2020 multi cloud is a big piece of the discussion and we're really happy to help unveil coming out of cell al kiram which is helping the networking challenges when it comes to multi cloud and I have the two co-founders they are brothers I have Amir who is the CEO and a DIF who is the CTO the Khan brothers thank you so much for joining us and congratulations on the launch of the company thank you sue for having us on the show it's a pleasure to see you again all right so Amir we've had you on the program your previous company that you've done was of course the fella you the two of you have worked together at I believe five companies successful companies acquired you know the most recent one into Cisco so a mirror obviously you know you know strong networking theme your brother the CTO I was going to talk to us about the engineering but give us you know just the the story of Al Kyra what you've been building and now ready to unveil to the world certainly needs to so in around 2018 timeframe we started looking into the next big problem to solve in the industry which was not only a substantial you know from the market size perspective but also from the customers perspective was solving a major pain point so when we started looking into the cloud customers and started talking to our customers they were struggling from the cloud networking perspective even in a single cloud and it was a new environment for them and they had to understand all the nitty-gritty details of each one of these clouds and when you go to multi cloud environment it becomes exponentially complicated to address not only connectivity but how to deploy services like firewall and other services including low balancers and IP address management etc and remote access so we started digging deeper into this problem and start working with the customers and took a clean sheet of paper and came up with a very comprehensive approach to offering a solution which is as a service this time we are not shipping any hardware or software it is you know just like any other SAS application you just come to our portal I just drag and drop literally draw out your network and click on provision and you know come back after 40 minutes or so your full global cloud infrastructure is up and running so out if your brother laid out a pretty broad vision there any of us from the networking world we know there's a lot of complexity there and therefore it takes a lot of work when I want to do things simply as a service is you know a huge growth area bring us inside the engineering challenges that you and the team have been working on to build this solution second let's do so we've been working both our men and myself in the networking industry for more than 25 years now and our the way we have worked and what we have believed in is that we need to solve customer problems we never believed in like doing a science project so here also we started working with customers as we have always done in the past we understood the customers pain points the challenges they were facing especially in this case and in cloud networking space multi-cloud networking space based on the user requirements users or the customers use cases we started the building a service and here what we have built is a complete network as a service it's a multi cloud met work as a service which not only provides connectivity to multiple routes but also addresses the needs for bringing in networking services as well as security services making sure that you have a full policy based infrastructure on top of it you have deep visibility into into the clouds as well as into on-premise into and visibility into and monitoring troubleshooting and all of it is delivered to you as a service so that's what we have been doing here at ELQ here excellent so when we look at multi-cloud of course you know every cloud they have some similar things they have some different things they all tend to do things a little bit differently you know one of the secret sauces that have been talked about for the last few years is ESP BAM space like you and built with Nutella to help really enable those environments so if we've got a diagram here which I think will help explain a little bit as you know we're out here it how it plugs into these different environments walk us through a little bit what we're seeing here and what you're actually doing a tell Kira so here we are building a global unifying the multi cloud Network it's consumed as a service think of it as consuming it just like you would consume any other SAS like our SAS issue so you come to lqs portal you register and then there you go and you start building your global multi-cloud unified network with integrated services so here what you see is is a Elka's cloud services exchange with comprises of cloud exchange points you can bring these up these cloud exchange points up anywhere on the globe you can decide like what networking services security services you need in these cloud exchange points you can connect the multiple clouds from there you can bring your existing on-premise connector matiee into the CX PS all these CX B's have a full mesh of overlay high speed low latency connectivity among each other so there is a full network which comes up between these CX B's and this the whole infrastructure scales with customers as as a customer scale so it's a horizontally scalable veil a very highly redundant and resilient infrastructure which we have both all right so armor now that we understand the basics of the technology you've got some strong investors including Sequoia kleiner perkins give us you know what is being announced day you're coming out of stealth where are you with the product you know how many employees you have and where are you with the discussion of customer adoption so stew we're obviously bringing this to the market and we will be announcing it on April 15th it's available for the customers to consume our solution as a service on that day so they are welcome to reach out to us and we'll be happy to help them and as a matter of fact just come to our website and register for the service and yeah we rightly said that we have a superstar team of not only the venture capital companies but also the board members representing those companies the bill Cochran and mamoon Hamid Wright who the leading VCS are on the board of our company including myself inactive all right I'm all right love to actually bring up the second slide that we have here walk us through you said you know the service you know how do people get started how do they understand you know what would walk us through what what they do so the biggest challenge when we started looking into these problems you know Stu was that it was very complicated you have to piecemeal bring up instances and the cloud and stitch them together and when you try to integrate the services that was a different challenge for the customers right so we wanted to make sure that it was so simple and clean that the customer didn't even have to think about any underlying construct on any of the clouds they should not have to worry about learning each individual power from the you know networking perspective so here's your portal you just come you know step one is come to a portal or register step two is you start drawing your network based on your intent what on-prem an activity you want to bring into this service what type of services you need like all all the firewalls and then you know what pilots you need to connect and everything happens seamlessly the from on pram pram through services into the cloud and across multiple clouds it's a seamless service that we have created and with full analytics capabilities and full governance built in alright so I'll to bring us into what this means for customers you know how do they manage it you know is this the networking team is it the cloud architects you know what api's are there how does this fit into kind of what customers are doing today and you know solve some of those challenges that we laid out earlier in the discussion yes trauma from the customers perspective it's as I said it's it's completely delivered as a service customers come to our portal they draw out the network they select the services they click on provision and the whole network comes up within minutes so the main thing here is that from a customer's point of view if they are connecting to different clouds they don't need to understand any of the underlying specifics or underlying constructs of any of the of the cloud in order to bring can I bring up connectivity so we what we are doing here is we are abstracting the clouds here so we are building a virtual cloud network so if you if you think of if you compare it with what we did in the in the previous life be virtualized the when so here would be a doing is we are virtualizing the cloud network so underlying doesn't matter which cloud you sit on which cloud you need to connect to which networking services whether cloud native services or whether you you want to consume our care services or we also support like customer bringing in third-party services as well so it's all all offered from our platform all offered is a service for to the customer again no expertise required in any of the underlying networking constructs of any of these cards give us what we should be looking at from a technology roadmap from Akira through the rest of 2020 good question as to so as I mentioned earlier our roadmap is dictated by customer requirements so we prioritize what customers need from us so we have come out with a scalable platform we have come out with a marketplace for networking services in there in the near term we'll be expanding our market place with more services we will be addressing more use cases and when I talk about use cases I can give you some examples like there's a view you not just only need connectivity into cloud you might have different requirements from from throughput perspective or bandwidth perspective or different services that you need to front-end your cloud but you may have certain applications such as internet basing application where you eat like traffic coming in from the internet inbound to those applications you might need services like a load balancer like an external load balancer in our services exchange you might also need like a firewall you might need traffic engineering or sorry service eaning capability is where you would chain service through multiple or traffic through multiple of these services like a firewall in the load balancer so we have built a platform which gives you all those capabilities going forward we will be adding more services more use cases to it we have a long ways ahead of us and we will be putting all our effort in delivering a roadmap as we go all right so Amma your technical team definitely has their hands full and uh you know robust after work on uh give us the the high-level what we should be looking for out Kira for people that are out there you know multi-cloud and networking you know tend to get talked a lot there's many big companies and some small ones what will separate al Kira from the rest of the market today and what should we be looking to see the company's progression through 2020 yeah thanks for asking that yeah certainly I mean you know from the solution perspective out it's said that you know it's so fundamentally important to have a very strong basis right and that's what we have done we are bringing out a certain number of services and now we will continue to grow on that will create a big marketplace we will continue to improve on which clouds we connect to and how and we will be building our own services in certain cases as well now building a technology is just one piece of it we have to go out to market with a company that the customers can trust every single you know the department in that company whether it's sales or how they do business with us all the business back-end pieces have to be sorted out and that's what we've been working with and you know then go to market partners that is very very important right support is very important so let me spend a little bit of time on go to market strategy we have been working with the service riders so that we can extend our reach not only to the large customers but also to midsize customers across the globe so you will see us in the future announcing major service water partnerships as well as we've been working with large sis bars and system integration in a partners and also we have taking a slightly different approach this time because it's a service so we are going with telecom master agents which have been you know working with the service providers the cloud providers the cable providers as a channel and they have a huge reach into the customer base so we we have a very comprehensive strategy not only from the go to market in the technology perspective but also how we are going to support our customers and continue to build a relationship to build a lasting company yeah I'm a super important point there absolutely we've seen the maturation and change in the service providers as today they are working with many of the public cloud providers and they're as you said a close touch point and a trusted partner of our customers all right so before I let you go you know YouTuber brothers everybody in today's day and age is spending even more time with family but you know your your situation you've worked together for a long time what keeps bringing the two of you together working together and then talk about that ball so I mean we're very close-knit family we have four brothers and one sister and obviously active and I have been the closest because we have been working together for the longest we have at least work in five different companies together our families travel together we have three daughters each we live about five minutes you know walk from each other and we you know just have this bond where we not only have you know the family close but also very close-knit friends a circle which we both hang out with and we you know obviously have common interest in the sports as well we play squash and tennis and work out so after four if they want to take a stab at it but also yeah so we've always been very close in fact we've been together for the last like ever since I can remember like even even college days he was we were roommates for for some time also he ever say we have like our circle of friends is the same also so again we're very close and we work well together so we complement each other's skills and and it's it's worked out in the past hopefully it will work out again and I look forward to working with them for many many more years to come yeah well I'm or not - thank you so much for sharing the the coming out of stealth for Al Kyra we definitely look forward to watching your progress and you know seeing how you're helping customers in this multi-cloud world thank you for joining us - thank you so much thank you for having us all right I'm Stu minimun and thank you so much for watching this special cube conversation on the cube [Music]

from the cube studios in Palo Alto in Boston connecting with thought leaders all around the world this is a cube conversation [Music] hello everybody and welcome to this week wiki bond cube insights powered by ETR in this breaking analysis we want to update you on the latest spending data from EGR as you know we've been tracking this weekly saga kodachi is here he's the director of research at ET our saga thanks for coming on thanks for having me again Dave really appreciate it yes so so let me remind everybody so we entered the Year this year 2020 with a consensus IT spend for cast of plus 4% once coronavirus hit ET are launched its latest survey in March and we saw those numbers you'll come down last week we reported well the first report we made was it looked like it was flat last week we reported a slight negative and today we want to update you guys on those numbers so saga before we get into the data just give us the high level on where you guys are at in terms of your survey yeah no problem so currently we are forecasting a decline in global IT budgets about negative 4% I think what's happened you know over the last you know 10 or 15 days is you've just seen more and more information released that's given organizations more of an understanding of just how severe this you know epidemic is and so what we've been able to do on our end is kind of do an event study analysis or simulation analysis kind of what you're seeing here a really pinpoint the time period where organizations understood the severity of the epidemic and then really trying to measure the declines in IT budgets from there great so guys bring that slide back up I want to share with our audience what's happening here so what ETR has done is an event-based analysis and what you can see is where the survey launched on 3/11 you could see how sentiment has declined literally daily as the data rolled in then you see the US declared a national emergency you saw that the federal plan leaked for that you know penned pandemic protect projection and obviously New York became a hot spot and then you can see this the stimulus package in it and sagger it looks like there's a slight uptick here but generally speaking it's down now it could be worse but you guys were the first to report the offset from work it worked from home infrastructure we'll talk about that a little bit talk about this event analysis and what you're seeing here and how you compressed the analysis hosting these events no problem so let's start with a blue line here and just so the audience knows the x-axis is going to be date and the y-axis is going to be annual growth or decline in nit budgets what you're seeing here and if we start with the blue line is we started pulling on 3/11 and on that date we started to ask you know fortune 100 is fortune 500 how their budget was going to change based on the impacts of coded nineteen versus their original expectations coming into coming into the year and again consensus estimates coming to the year were positive four percent so if you track that line all the way through you get to a decline of about one percent now what's the issue of starting polling on 3/11 or using that blue line well one of the big issues is a few days later the US declared a national emergency so more information was released right I think organizations that took the survey in the first two days didn't have a complete picture as to what's going on and then effectively a week later you saw federal documents get leaked stating how bad this epidemic was right in terms of the last 18 18 plus months and so what we did was we did it effectively an event based analysis or defuse different simulation where if you take a look at the yellow and red lines to start what we're doing is we're effectively saying okay let's ignore everyone that took the survey prior to that let's take their budgets in terms of how they indicated change versus their original expectations for 2020 and then let's go ahead and map that and if you look at the yellow line as an example that goes to a decline of 2% and then once I think you know the next shoe dropped in terms of organizations understanding this is not going to be a few weeks or this is not the common cold or flu once organizations knew this was going to be an 18 plus epidemic you can see if we started pulling respondents from there how much more negative it gets and of course once NYC became the epicenter you saw a little another shoe drop so now those those scenarios or simulations are taking us between a decline of three and four percent and then of course if we look at that last purple line there when the stimulus got announced what we are seeing is it looks like it may have bottomed down we have to continue tracking it because you know again it's just a few days since the stimulus is was passed and so let's see if the data starts improve a little bit or at least stabilize but I think from the last three events in terms of the the federal plan being leaked NYC becoming the epicenter and the stimulus it looks like the market now is fully aware of what's going on and now we're kind of seeing some stabilization in the data in terms of the declines for 2020 so between the feds action and the the fiscal stimulus we've we've seen some optimism although people are really cautious of course remember folks this would be worse were it not for the shift in spend to work from home infrastructure not just collaboration and visualization tools but other infrastructure around that network bandwidth security desktop virtualization etc so guys if you bring up the next chart I want to set this up we've been reporting this framework for a while now what this shows is what the sentiment is in terms of the budget change and you can see the gray bar now is 35% it started at 40% so that's dropped so the percentage of CIO saying no change the green is held pretty steady at around 20 to 22% that's it's roughly in there and the red you know has been has been shifting and you can see most of the green ie spending more in 2020 is focused on that you know one to two ten percent but but Sagar bring us up to date now we're going to settle in it right now about three and a half to four percent on the negative side give us some color on this chart please yeah no problem so the best way to connect this chart with what we saw earlier is this is a snapshot so this is a single day so this is the data that is feeding the time series chart kind of help the audience understand what's going on so if we were to look at this exact chart Oh since March 11 you would see that midpoint Average effectively coming down every day and that's effectively what's making up that time series in terms of this chart you know Dave you kind of hit it right on the nail you're kind of seeing the positivity remain or be stable and again that's that work from home infrastructure as you as you mentioned right the collaboration pools no the virtualization support services networking bandwidth all that stuff right being more and more security but on the negative side I think what you're seeing is that again as organizations now understand the severity of the epidemic I think as we understand further and we've talked about this you know a few weeks ago that organizations were anticipating less demand they were anticipating an uptick in broken supply chains now you're starting to see some of that play out and as a result you're seeing organizations get more and more negative and that's why that midpoint average it keeps declining that's why those red bars keep going up is the the impacts in you know based on the data are are now starting to be to be seen and so you know let's see if the stimulus stabilizes this data and we'll continue tracking that you know over the next few weeks the next few months okay so basically we're coming in - three and a half to four percent that's where we are today we're not going to get detailed into some of the vendors today we talked a little bit about that last week and go back to last week's breaking analysis you can see some of that vendor commentary I want to talk about what happens next ETR now we'll go into a two-week quite self-imposed quiet period and really start crunching the data at the end of that quiet period they will release to their private clients the their latest thinking in a webcast after that time we at the cube are allowed to share public information and we're gonna drill down into some of the segments that our community is most interested in but-but-but etrs going quiet now so saga maybe you can explain that sequence and fill in any holes that I missed there yeah no problem the next two weeks so we've we've collected a tremendous amount of data you know we're over you know we're at a hundred fortune 100 organizations you know almost three four hundred global two thousand organizations and so we're at a point now where it's time to start aggregating the data start really analyzing it going through this Koga drill down that we conducted but also we conducted a tremendous study on technology spending intentions of crossing over 350 vendors dozens of Technology sectors and so now it's really a time to kind of drill in and you know what what we're looking for or even some of the biggest takeaways from from this Cove it you know drill down is you know if if you started polling before 3:23 chances are your forecast is gonna come in light and I think that's one of the things that we've learned as we're kind of going into this to hear it is we really want to measure the impact starting right around that 3:23 timeframe it looks right around then based on that time series chart that we showed earlier that's when the market fully understood the impact of this epidemic and so as we start over the next two weeks even though we started pulling a little bit early we really want to focus on that second set a second half of responses because that's probably gonna be more indicative of what's going on I think the second thing is gonna be look if condition of conditions continue to deteriorate things can get worse and so we may come out of the next two weeks with this data that we collected and again have to continue indicating that you know the environment has continued coming down and you know maybe we may have to make adjustments as we see fit so I think that's kind of you know this whole situation is so dynamic still and so we're gonna do our best in the next week and a half to kind of get this data to market to at least give everyone an idea here's how everything stands right now and so that people have a good benchmark and then move forward yeah so this is as close to real time really as you can get in some of this IT spending world saga mentioned some of the numbers and in the global 2000 fortune fortune 100 1000 this this end now just the reminder is up over 1200 I believe right Sahra the total and that you've collected this this month that's correct exactly every time we've been doing one of these it's been going up another a couple hundred respondents so yeah we're at a very comfortable level now our sample right now represents five hundred and fifty five billion dollars in annual IP spend you know and global IT spend every year is a little over you know three trillion so this is a significant significant portion of a global IT spend and we feel comfortable at this point kind of going into that quiet period as you mentioned and really start to dig through the results that you know now that we've kind of you know covered the the 10,000 foot or the macro layer so to speak in terms of where budgets are going now it's really time to start drilling down and do the sectors and vendors because this is this is not going to be a every vendors going down or whatever maybe there's so many different dynamics here some vendors are going to do very well because the work for MoMA infrastructure and I think some vendors are gonna do very poorly because one they're not only on the legacy side but they're not really aligned from this whole work from home infrastructure movement so you're gonna see a lot of bifurcation you know as we get into 53 that's right and we're gonna dig into all those segments we're gonna look at the work from home we're gonna look at the traditional stuff we're gonna look at cloud we're gonna drill into specific segments that are that are of interest to our community it's a pleasure to really have you on here Sagar thank you for for sharing giving us access to this data and and stay safe and we will be watching go to ETR dot plus and you know check out what's happening there Silicon Engel Tom will obviously cover this and I published weekly on wiki bond comm again that saga thanks so much for coming on the cube yeah no problem thank you so much and looking forward to catching up in a few weeks all right then thank you for watching everybody this is Dave a latte for the cube or wiki bounce cube insights powered by ETR we'll see you next time [Music]

hello everybody and thank you for joining us today at the virtual Vertica BBC 2020 today's breakout session is entitled the road to autonomous database management how Domo is delivering SLA for less my name is su LeClair I'm the director of marketing at Vertica and I'll be your host for this webinar joining me is Ben white senior database engineer at Domo but before we begin I want to encourage you to submit questions or comments during the virtual session you don't have to wait just type your question or comment in the question box below the slides and click Submit there will be a Q&A session at the end of the presentation we'll answer as many questions as we're able to during that time any questions that we aren't able to address or drew our best to answer them offline alternatively you can visit vertical forums to post your questions there after the session our engineering team is planning to join the forum to keep the conversation going also as a reminder you can maximize your screen by clicking the double arrow button in the lower right corner of the slide and yes this virtual session is being recorded and will be available to view on demand this week we'll send you notification as soon as it's ready now let's get started then over to you greetings everyone and welcome to our virtual Vertica Big Data conference 2020 had we been in Boston the song you would have heard playing in the intro would have been Boogie Nights by heatwaves if you've never heard of it it's a great song to fully appreciate that song the way I do you have to believe that I am a genuine database whisperer then you have to picture me at 3 a.m. on my laptop tailing a vertical log getting myself all psyched up now as cool as they may sound 3 a.m. boogie nights are not sustainable they don't scale in fact today's discussion is really all about how Domo engineers the end of 3 a.m. boogie nights again well I am Ben white senior database engineer at Domo and as we heard the topic today the road to autonomous database management how Domo is delivering SLA for less the title is a mouthful in retrospect I probably could have come up with something snazzy er but it is I think honest for me the most honest word in that title is Road when I hear that word it evokes for me thoughts of the journey and how important it is to just enjoy it when you truly embrace the journey often you look up and wonder how did we get here where are we and of course what's next right now I don't intend to come across this too deep so I'll submit there's nothing particularly prescient and simply noticing the elephant in the room when it comes to database economy my opinion is then merely and perhaps more accurately my observation the office context imagine a place where thousands and thousands of users submit millions of ad-hoc queries every hour now imagine someone promised all these users that we could deliver bi leverage at cloud scale in record time I know what many of you should be thinking who in the world would do such a thing of course that news was well received and after the cheers from executives and business analysts everywhere and chance of Keep Calm and query on finally started to subside someone that turns an ass that's possible we can do that right except this is no imaginary place this is a very real challenge we face the demo through imaginative engineering demo continues to redefine what's possible the beautiful minds at Domo truly embrace the database engineering paradigm that one size does not fit all that little philosophical nugget is one I would pick up while reading the white papers and books of some guy named stone breaker so to understand how I and by extension Domo came to truly value analytic database administration look no further than that philosophy and what embracing it would mean it meant really that while others were engineering skyscrapers we would endeavor to build Datta neighborhoods with a diverse kapala G of database configuration this is where our journey at Domo really gets under way without any purposeful intent to define our destination not necessarily thinking about database as a service or anything like that we had planned this ecosystem of clusters capable of efficiently performing varied workloads we achieve this with custom configurations for node count resource pool configuration parameters etc but it also meant concerning ourselves with the unattended consequences of our ambition the impact of increased DDL activities on the catalog system overhead in general what would be the management requirements of an ever-evolving infrastructure we would be introducing multiple points of failure what are the advantages the disadvantages those types of discussions and considerations really help to define what would be the basic characteristics of our system the database itself needed to be trivial redundant potentially ephemeral customizable and above all scalable and we'll get more into that later with this knowledge of what we were getting into automation would have to be an integral part of development one might even say automation will become the first point of interest on our journey now using popular DevOps tools like saltstack terraform ServiceNow everything would be automated I mean it discluded everything from larger multi-step tasks like database designs database cluster creation and reboots to smaller routine tasks like license updates move-out and projection refreshes all of this cool automation certainly made it easier for us to respond to problems within the ecosystem these methods alone still if our database administration reactionary and reacting to an unpredictable stream of slow query complaints is not a good way to manage a database in fact that's exactly how three a.m. Boogie Nights happen and again I understand there was a certain appeal to them but ultimately managing that level of instability is not sustainable earlier I mentioned an elephant in the room which brings us to the second point of interest on our road to autonomy analytics more specifically analytic database administration why our analytics so important not just in this case but generally speaking I mean we have a whole conference set up to discuss it domo itself is self-service analytics the answer is curiosity analytics is the method in which we feed the insatiable human curiosity and that really is the impetus for analytic database administration analytics is also the part of the road I like to think of as a bridge the bridge if you will from automation to autonomy and with that in mind I say to you my fellow engineers developers administrators that as conductors of the symphony of data we call analytics we have proven to be capable producers of analytic capacity you take pride in that and rightfully so the challenge now is to become more conscientious consumers in some way shape or form many of you already employ some level of analytics to inform your decisions far too often we are using data that would be categorized as nagging perhaps you're monitoring slow queries in the management console better still maybe you consult the workflows analyzing how about a logging and alerting system like sumo logic if you're lucky you do have demo where you monitor and alert on query metrics like this all examples of analytics that help inform our decisions being a Domo the incorporation of analytics into database administration is very organic in other words pretty much company mandated as a company that provides BI leverage a cloud scale it makes sense that we would want to use our own product could be better at the business of doma adoption of stretches across the entire company and everyone uses demo to deliver insights into the hands of the people that need it when they need it most so it should come as no surprise that we have from the very beginning use our own product to make informed decisions as it relates to the application back engine in engineering we call it our internal system demo for Domo Domo for Domo in its current iteration uses a rules-based engine with elements through machine learning to identify and eliminate conditions that cause slow query performance pulling data from a number of sources including our own we could identify all sorts of issues like global query performance actual query count success rate for instance as a function of query count and of course environment timeout errors this was a foundation right this recognition that we should be using analytics to be better conductors of curiosity these types of real-time alerts were a legitimate step in the right direction for the engineering team though we saw ourselves in an interesting position as far as demo for demo we started exploring the dynamics of using the platform to not only monitor an alert of course but to also triage and remediate just how much economy could we give the application what were the pros and cons of that Trust is a big part of that equation trust in the decision-making process trust that we can mitigate any negative impacts and Trust in the very data itself still much of the data comes from systems that interacted directly and in some cases in directly with the database by its very nature much of the data was past tense and limited you know things that had already happened without any reference or correlation to the condition the mayor to those events fortunately the vertical platform holds a tremendous amount of information about the transaction it had performed its configurations the characteristics of its objects like tables projections containers resource pools etc this treasure trove of metadata is collected in the vertical system tables and the appropriately named data collector tables as a version 9 3 there are over 190 tables that define the system tables while the data collector is the collection of 215 components a rich collection can be found in the vertical system tables these tables provide a robust stable set of views that let you monitor information about your system resources background processes workload and performance allowing you to more efficiently profile diagnose and correlate historical data such as low streams query profiles to pool mover operations and more here you see a simple query to retrieve the names and descriptions of the system tables and an example of some of the tables you'll find the system tables are divided into two schemas the catalog schema contains information about persistent objects and the monitor schema tracks transient system States most of the tables you find there can be grouped into the following areas system information system resources background processes and workload and performance the Vertica data collector extends system table functionality by gathering and retaining aggregating information about your database collecting the data collector mixes information available in system table a moment ago I show you how you get a list of the system tables in their description but here we see how to get that information for the data collector tables with data from the data collecting tables in the system tables we now have enough data to analyze that we would describe as conditional or leading data that will allow us to be proactive in our system management this is a big deal for Domo and particularly Domo for demo because from here we took the critical next step where we analyze this data for conditions we know or suspect lead to poor performance and then we can suggest the recommended remediation really for the first time we were using conditional data to be proactive in a database management in record time we track many of the same conditions the Vertica support analyzes via scrutinize like tables with too many production or non partition fact tables which can negatively affect query performance and life in vertical in viral suggests if the table has a data a time step column you recommend the partitioning by the month we also can track catalog sizes percentage of total memory and alert thresholds and trigger remediations requests per hour is a very important metric in determining when a trigger are scaling solution tracking memory usage over time allows us to adjust resource pool parameters to achieve the optimal performance for the workload of course the workload analyzer is a great example of analytic database administration I mean from here one can easily see the logical next step where we were able to execute these recommendations manually or automatically be of some configuration parameter now when I started preparing for this discussion this slide made a lot of sense as far as the logical next iteration for the workload analyzing now I left it in because together with the next slide it really illustrates how firmly Vertica has its finger on the pulse of the database engineering community in 10 that OS management console tada we have the updated work lies will load analyzer we've added a column to show tuning commands the management console allows the user to select to run certain recommendations currently tuning commands that are louder and alive statistics but you can see where this is going for us using Domo with our vertical connector we were able to then pull the metadata from all of our clusters we constantly analyze that data for any number of known conditions we build these recommendations into script that we can then execute immediately the actions or we can save it to a later time for manual execution and as you would expect those actions are triggered by thresholds that we can set from the moment nyan mode was released to beta our team began working on a serviceable auto-scaling solution the elastic nature of AI mode separated store that compute clearly lent itself to our ecosystems requirement for scalability in building our system we worked hard to overcome many of the obstacles they came with the more rigid architecture of enterprise mode but with the introduction is CRM mode we now have a practical way of giving our ecosystem at Domo the architectural elasticity our model requires using analytics we can now scale our environment to match demand what we've built is a system that scales without adding management overhead or our necessary cost all the while maintaining optimal performance well we're really this is just our journey up to now and which begs the question what's next for us we expand the use of Domo for Domo within our own application stack maybe more importantly we continue to build logic into the tools we have by bringing machine learning and artificial intelligence to our analysis and decision making really do to further illustrate those priorities we announced the support for Amazon sage maker autopilot at our demo collusive conference just a couple of weeks ago for vertical the future must include in database economy the enhanced capabilities in the new management console to me are clear nod to that future in fact with a streamline and lightweight database design process all the pieces should be in place versions deliver economists database management itself we'll see well I would like to thank you for listening and now of course we will have a Q&A session hopefully very robust thank you [Applause]

>> Jeff: Hello everybody and thank you for joining us today for the Virtual VERTICA BDC 2020. Today's breakout session has been titled, "VERTICA Database Designer Today and Tomorrow." I'm Jeff Healey, Product VERTICA Marketing, I'll be your host for this breakout session. Joining me today is Yuanzhe Bei, Senior Technical Manager from VERTICA Engineering. But before we begin, (clearing throat) I encourage you to submit questions or comments during the virtual session. You don't have to wait, just type your question or comment in the question box below the slides and click Submit. As always, there will be a Q&A session at the end of the presentation. We'll answer as many questions, as we're able to during that time, any questions we don't address, we'll do our best to answer them offline. Alternatively, visit VERTICA forums at forum.vertica.com to post your questions there after the session. Our engineering team is planning to join the forums, to keep the conversation going. Also, a reminder that you can maximize your screen by clicking the double arrow button at the lower right corner of the slides. And yes, this virtual session is being recorded and will be available to view on demand this week. We will send you a notification as soon as it's ready. Now let's get started. Over to you Yuanzhe. >> Yuanzhe: Thanks Jeff. Hi everyone, my name is Yuanzhe Bei, I'm a Senior Technical Manager at VERTICA Server RND Group. I run the query optimizer, catalog and the disaggregated engine team. Very glad to be here today, to talk about, the "VERTICA Database Designer Today and Tomorrow". This presentation will be organized as the following; I will first refresh some knowledge about, VERTICA fundamentals such as Tables and Projections, which will bring to the question, "What is Database Designer?" and "Why we need this tool?". Then I will take you through a deep dive, into a Database Designer or we call DBD, and see how DBD's internals works, after that I'll show you some exciting DBD improvements, we have planned for 10.0 release and lastly, I will share with you, some DBD future roadmap we planned next. As most of you should already know, VERTICA is built on a columnar architecture. That means, data is stored column wise. Here we can see a very simple example, of table with four columns, and the many of you may also know, table in VERTICA is a virtual concept. It's just a logical representation of data, which means user can write SQL query, to reference the table names and column, just like other relational database management system, but the actual physical storage of data, is called Projection. A Projection can reference a subset, or all of the columns all to its anchor table, and must be sorted by at least one column. Each table need at least one C for projection which reference all the columns to the table. If you load data to a table with no projection, and automated, auto production will be created, which will be arbitrarily assorted by, the first couple of columns in the table. As you can imagine, even though such other production, can be used to answer any query, the performance is not optimized in most cases. A common practice in VERTICA, is to create multiple projections, contain difference step of column, and sorted in different ways on the same table. When query is sent to the server, the optimizer will pick the projection, that can answer the query in the most efficient way. For example, here you can say, let's say you have a query, that select columns B, D, C and sorted by B and D, the third projection will be ideal, because the data is already sorted, so you can save the sorting costs while executing the query. Basically when you choose the design of the projection, you need to consider four things. First and foremost, of course the sort order. The data already sorted in the right way, can benefit quite a lot of the query actually, like Ordered by, Group By, Analytics, Merge, Join, Predicates and so on. The select column group is also important, because the projection must contain, all the columns referenced by your workflow query. Even missing one column in the projection, this projection cannot be used for a particular query. In addition, VERTICA is the distributed database, and allow projection to be segmented, based on the hash of a set of columns, which is beneficial if the segmentation merged, the join keys or group keys. And finally encoding of each per columns is also part of the design, because the data is sorted in different way, may completely change the optimal encoding for each column. This example only show the benefit of the first two, but you can imagine the rest too are also important. But even for that, it doesn't sound that hard, right? Well I hope you change your mind already when you see this, at least I do. These machine generated queries, really beats me. It will probably take an experienced DBA hours, to figure out which projection can be benefit these queries, not even mentioning there could be hundreds of such queries, in the regular work logs in the real world. So what can we do? That's why we need DBD. DBD is a tool integrated in the VERTICA server, that it can help DBA to perform an access, on their work log query, tabled schema and data, and then automatically figure out, the most optimized projection design for their workload. In addition, DBD also a sophisticated tool, that can take customize by a user, by sending a lot of parameters objectives and so on. And lastly, DBD has access to the optimizer, so DB knows what kind of attribute, the projection need to have, in order to have the optimizer to benefit from them. DBD has been there for years, and I'm sure there are plenty of materials available online, to show you how DBD can be used in different scenarios, whether to achieve the query optimize, or load optimize, whether it's the comprehensive design, or the incremental design, whether it's a dumping deployment script, and manual deployment later, or let the DBD do the order deployment for you, and the many other options. I'm not planning to talk about this today, instead, I will take the opportunity today, to open this black box DBD, and show you what exactly hide inside. DBD is a complex tool and I have tried my best to summarize the DBD design process into seven steps; Extract, Permute, Prune, Build, Score, Identify and Encode. What do they mean? Don't worry, I will show you step by step. The first step is Extract. Extract Interesting Columns. In this step, DBD pass the design queries, and figure out the operations that can be benefited, by the potential projection design, and extract the corresponding columns, as interesting columns. So Predicates, Group By, Order By, Joint Condition, and analytics are all interesting Column to the DBD. As you can see this three simple sample queries, DBD can extract the interest in column sets on the right. Some of these column sets are unordered. For example, the green one for Group By a1 and b1, the DBD extracts the interesting column set, and put them in the own orders set, because either data sorted by a1 first or b1 first, can benefit from this Group By operation. Some of the other sets are ordered, and the best example is here, order by clause a2 and b2, and obviously you cannot sort it by b2 and then a2. These interesting columns set will be used as if, to extend to actual projection sort order candidates. The next step is Permute, once DBD extract all the C's, it will enumerate sort order using C, and how does DBD do that? I'm starting with a very simple example. So here you can see DBD can enumerate two sort orders, by extending d1 with the unordered set a1, b1, and the derived at two sort order candidates, d1, a1, b1, and d1, b1, a1. This sort order can benefit queries with predicate on d1, and also benefit queries by Group By a1, b1, when a1, sorry when d1 is constant. So with the same idea, DBD will try to extend other States with each other, and populate more sort order permutations. You can imagine that how many of them, there could be many of them, these candidates, based on how many queries you have in the design and that can be handled of the sort order candidates. That comes to the third step, which is Pruning. This step is to limit the candidates sort order, so that the design won't be running forever. DBD uses very simple capping mechanism. It sorts all the, sort all the candidates, are ranked by length, and only a certain number of the sort order, with longest length, will be moved forward to the next step. And now we have all the sort orders candidate, that we want to try, but whether this sort order candidate, will be actually be benefit from the optimizer, DBD need to ask the optiizer. So this step before that happens, this step has to build those projection candidate, in the catalog. So this step will build, will generates the projection DBL's, surround the sort order, and create this projection in the catalog. These projections won't be loaded with real data, because that takes a lot of time, instead, DBD will copy over the statistic, on existing projections, to this projection candidates, so that the optimizer can use them. The next step is Score. Scoring with optimizer. Now projection candidates are built in the catalog. DBD can send a work log queries to optimizer, to generate a query plan. And then optimizer will return the query plan, DBD will go through the query plan, and investigate whether, there are certain benefits being achieved. The benefits list have been growing over time, when optimizer add more optimizations. Let's say in this case because the projection candidates, can be sorted by the b1 and a1, it is eligible for Group By Pipe benefit. Each benefit has a preset score. The overall benefit score of all design queries, will be aggregated and then recorded, for each projection candidate. We are almost there. Now we have all the total benefit score, for the projection candidates, we derived on the work log queries. Now the job is easy. You can just pick the sort order with the highest score as the winner. Here we have the winner d1, b1 and a1. Sometimes you need to find more winners, because the chosen winner may only benefit a subset, of the work log query you provided to the DBD. So in order to have the rest of the queries, to be also benefit, you need more projections. So in this case, DBD will go to the next iteration, and let's say in this case find to another winner, d1, c1, to benefit the work log queries, that cannot be benefit by d1, b1 and a1. The number of iterations and thus the winner outcome, DBD really depends on the design objective that uses that. It can be load optimized, which means that only one, super projection winner will be selected, or query optimized, where DBD try to create as many projections, to cover most of the work log queries, or somewhat balance an objective in the middle. The last step is to decide encoding, for each projection columns, for the projection winners. Because the data are sorted differently, the encoding benefits, can be very different from the existing projection. So choose the right projection encoding design, will save the disk footprint a significant factor. So it's worth the effort, to find out the best thing encoding. DBD picks the encoding, based on the actual sampling the data, and measure the storage footprint. For example, in this case, the projection winner has three columns, and say each column has a few encoding options. DBD will write the sample data in the way this projection is sorted, and then you can see with different encoding, the disk footprint is different. DBD will then compare the disk footprint of each, of different options for each column, and pick the best encoding options, based on the one that has the smallest storage footprint. Nothing magical here, but it just works pretty well. And basic that how DBD internal works, of course, I think we've heard it quite a lot. For example, I didn't mention how the DBD handles segmentation, but the idea is similar to analyze the sort order. But I hope this section gave you some basic idea, about DBD for today. So now let's talk about tomorrow. And here comes the exciting part. In version 10.0, we significantly improve the DBD in many ways. In this talk I will highlight four issues in old DBD and describe how the 10.0 version new DBD, will address those issues. The first issue is that a DBD API is too complex. In most situations, what user really want is very simple. My queries were slow yesterday, with the new or different projection can help speed it up? However, to answer a simple question like this using DBD, user will be very likely to have the documentation open on the side, because they have to go through it's whole complex flow, from creating a projection, run the design, get outputs and then create a design in the end. And that's not there yet, for each step, there are several functions user need to call in order. So adding these up, user need to write the quite long script with dozens of functions, it's just too complicated, and most of you may find it annoying. They either manually tune the projection to themselves, or simply live with the performance and come back, when it gets really slow again, and of course in most situations, they never come back to use the DBD. In 10.0 VERTICA support the new simplified API, to run DBD easily. There will be just one function designer_single_run and one argument, the interval that you think, your query was slow. In this case, user complained about it yesterday. So what does this user to need to do, is just specify one day, as argument and run it. The user don't need to provide anything else, because the DBD will look up his query or history, within that time window and automatically populate design, run design and export the projection design, and the clean up, no user intervention needed. No need to have the documentation on the side and carefully write a script, and a debug, just one function call. That's it. Very simple. So that must be pretty impressive, right? So now here comes to another issue. To fully utilize this single round function, users are encouraged to run DBD on the production cluster. However, in fact, VERTICA used to not recommend, to run a design on a production cluster. One of the reasons issue, is that DBD picks massive locks, both table locks and catalog locks, which will badly interfere the running workload, on a production cluster. As of 10.0, we eliminated all the table and ten catalog locks from DBD. Yes, we eliminate 100% of them, simple improvement, clear win. The third issue, which user may not be aware of, is that DBD writes intermediate result. into real VERTICA tables, the real DBD have to do that is, DBD is the background task. So the intermediate results, some user needs to monitor it, the progress of the DBD in concurrent session. For complex design, the intermediate result can be quite massive, and as a result, many lost files will be created, and written to the disk, and we should both stress, the catalog, and that the disk can slow down the design. For ER mode, it's even worse because, the table are shared on communal storage. So writing to the regular table, means that it has to upload the data, to the communal storage, which is even more expensive and disruptive. In 10.0, we significantly restructure the intermediate results buffer, and make this shared in memory data structure. Monitoring queries will go directly look up, in memory data structure, and go through the system table, and return the results. No Intermediate Results files will be written anymore. Another expensive lubidge of local disk for DBD is encoding design, as I mentioned earlier in the deep dive, to determine which encoding works the best for the new projection design, there's no magic way, but the DBD need to actually write down, the sample data to the disk, using the different encoding options, and to find out which ones have the smallest footprint, or pick it as the best choice. These written sample data will be useless after this, and it will be wiped out right away, and you can imagine this is a huge waste of the system resource. In 10.0 we improve this process. So instead of writing, the different encoded data on the disk, and then read the file size, DBD aggregate the data block size on-the-fly. The data block will not be written to the disk, so the overall encoding and design is more efficient and non-disruptive. Of course, this is just about the start. The reason why we put a significant amount of the resource on the improving the DBD in 10.0, is because the VERTICA DBD, as essential component of the out of box performance design campaign. To simply illustrate the timeline, we are now on the second step, where we significantly reduced, the running overhead of the DBD, so that user will no longer fear, to run DBD on their production cluster. Please be noted that as of 10.0, we haven't really started changing, how DBD design algorithm works, so that what we have discussed in the deep dive today, still holds. For the next phase of DBD, we will briefly make the design process smarter, and this will include better enumeration mechanism, so that the pruning is more intelligence rather than brutal, then that will result in better design quality, and also faster design. The longer term is to make DBD to achieve the automation. What entail automation and what I really mean is that, instead of having user to decide when to use DBD, until their query is slow, VERTICA have to know, detect this event, and have have DBD run automatically for users, and suggest the better projections design, if the existing projection is not good enough. Of course, there will be a lot of work that need to be done, before we can actually fully achieve the automation. But we are working on that. At the end of day, what the user really wants, is the fast database, right? And thank you for listening to my presentation. so I hope you find it useful. Now let's get ready for the Q&A.

>> Jeff: Hello everybody, and thank you for joining us today, in the virtual Vertica BDC 2020. Today's Breakout session is entitled, "Sizing and Configuring Vertica in Eon Mode for Different Use Cases". I'm Jeff Healey, and I lead Vertica Marketing. I'll be your host for this Breakout session. Joining me are Sumeet Keswani, and Shirang Kamat, Vertica Product Technology Engineers, and key leads on the Vertica customer success needs. But before we begin, I encourage you to submit questions or comments during the virtual session, you don't have to wait, just type your question or comment in the question box below the slides, and click submit. There will be a Q&A session at the end of the presentation, we will answer as many questions as we're able to during that time, any questions we don't address, we'll do our best to answer them off-line. Alternatively, visit Vertica Forums, at forum.vertica.com, post your question there after the session. Our Engineering Team is planning to join the forums to keep the conversation going. Also as reminder, that you can maximize your screen by clicking the double arrow button in the lower-right corner of the slides, and yes, this virtual session is being recorded, and will be available to view on-demand this week. We'll send you a notification as soon as it's ready. Now let's get started! Over to you, Shirang. >> Shirang: Thanks Jeff. So, for today's presentation, we have picked Eon Mode concepts, we are going to go over sizing guidelines for Eon Mode, some of the use cases that you can benefit from using Eon Mode. And at last, we are going to talk about, some tips and tricks that can help you configure and manage your cluster. Okay. So, as you know, Vertica has two modes of operation, Eon Mode and Enterprise Mode. So the question that you may have is, which mode should I implement? So let's look at what's there in the Enterprise Mode. Enterprise Mode, you have a cluster, with general purpose compute nodes, that have locally at their storage. Because of this tight integration of compute and storage, you get fast and reliable performance all the time. Now, amount of data that you can store in Enterprise Mode cluster, depends on the total disk capacity of the cluster. Again, Enterprise Mode is more suitable for on premise and cloud deployments. Now, let's look at Eon Mode. To take advantage of cloud economics, Vertica implemented Eon Mode, which is getting very popular among our customers. In Eon Mode, we have compute and storage, that are separated by introducing S3 Bucket, or, S3 compliant storage. Now because of this separation of compute and storage, you can take advantages like mapping all dynamic scale-out and scale-in. Isolation of your workload, as well as you can load data in your cluster, without having to worry about the total disk capacity of your local nodes. Obviously, you know, it's obvious from what they accept, Eon Mode is suitable for cloud deployment. Some of our customers who take advantage of the features of Eon Mode, are also deploying it on premise, by introducing S3 compliant slash web storage. Okay? So, let's look at some of the terminologies used in Eon Mode. The four things that I want to talk about are, communal storage. It's a shared storage, or S3 compliant shared storage, a bucket that is accessible from all the nodes in your cluster. Shard, is a segment of data, stored on the communal storage. Subscription, is the binding with nodes and shards. And last, depot. Depot is a local copy or, a local cache, that can help query in group performance. So, shard is a segment of data stored in communal storage. When you create a Eon Mode cluster, you have to specify the shard count. Shard count decide the maximum number of nodes that will participate in your query. So, Vertica also will introduce a shard, called replica shard, that will hold the data for replicated projections. Subscriptions, as I said before, is a binding between nodes and shards. Each node subscribes to one or more shards, and a shard has at least two nodes that subscribe to it for case 50. Subscribing nodes are responsible for writing and reading from shard data. Also subscriber node holds up-to-date metadata for a catalog of files that are present in the shard. So, when you connect to Vertica node, Vertica will automatically assign you set of nodes and subscriptions that will process your query. There are two important system tables. There are node subscriptions, and session subscriptions, that can help you understand this a little bit more. So let's look at what's on the local disk of your Eon Mode cluster. So, on local disk, you have depot. Depot is a local file system cache, that can hold subset of the data, or copy of the data, in communal storage. Other things that are there, are temp storage, temp storage is used for storing data belonging to temporary tables, and, the data that spills through this, when you are processing queries. And last, is catalog. Catalog is a persistent copy of Vertica, catalog that is written to this. The writes happen at every commit. You only need the persistent copy at node startup. There is also a copy of Vertica catalog, stored in communal storage, called durability. The local copy is synced to the copy in communal storage via service, at the interval of five minutes. So, let's look at depot. Now, as I said before, depot is your file system cache. It's help to reduce network traffic, and slow performance of your queries. So, we make assumption, that when we load data in Vertica, that's the data that you may most frequently query. So, every data that is loaded in Vertica is first entering the depot, and then as a part of same transaction, also synced to communal storage for durability. So, when you query, when you run a query against Vertica, your queries are also going to find the files in the depot first, to be used, and if the files are not found, the queries will access files from communal storage. Now, the behavior of... you know, the new files, should first enter the depot or skip depot can be changed by configuration parameters that can help you skip depot when writing. When the files are not found in depot, we make assumption that you may need those files for future runs of your query. Which means we will fetch them asynchronously into the depot, so that you have those files for future runs. If that's not the behavior that you intend, you can change configuration around return, to tell Vertica to not fetch them when you run your query, and this configuration parameter can be set at database level, session level, query level, and we are also introducing a user level parameter, where you can change this behavior. Because the depot is going to be limited in size, compared to amount of data that you may store in your Eon cluster, at some point in time, your depot will be full, or hit the capacity. To make space for new data that is coming in, Vertica will evict some of the files that are least frequently used. Hence, depot is going to be your query performance enhancer. You want to shape the extent of your depot. And, so what you want to do is, to decide what shall be in your depot. Now Vertica provides some of the policies, called pinning policies, that can help you pin of statistics table or addition of a table, into a depot, at subcluster level, or at the database level. And Sumeet will talk about this a bit more in his future slides. Now look at some of the system tables that can help you understand about the size of the depot, what's in your depot, what files were evicted, what files were recently fetched into the depot. One of the important system tables that I have listed here is DC_FILE_READS. DC_FILE_READS can be used to figure out if your transaction or query fetched with data from depot, from communal storage, or component. One of the important features of Eon Mode is a subcluster. Vertica lets you divide your cluster into smaller execution groups. Now, each of the execution groups has a set of nodes together subscribed to all the shards, and can process your query independently. So when you connect one node in the subcluster, that node, along with other nodes in the subcluster, will only process your query. And because of that, we can achieve isolation as well as, you know, fetches, scale-out and scale-in without impacting what's happening on the cluster. The good thing about subclusters, is all the subclusters have access to the communal storage. And because of this, if you load data in one subcluster, it's accessible to the queries that are running in other subclusters. When we introduced subclusters, we knew that our customers would really love these features, and, some of the things that we were considering is, we knew that our customers would dynamically scale out and in, lots of-- they would add and remove lots of subclusters on demand, and we had to provide that ab-- we had to give this feature, or provide ability to add and remove subclusters in a fast and reliable way. We knew that during off-peak hours, our customers would shut down many of their subclusters, that means, more than half of the nodes could be down. And we had to make adjustment to our quorum policy which requires at least half of the nodes to be up for database to stay up. We also were aware that customers would add hundreds of nodes in the cluster, which means we had to make adjustments to the catalog and commit policy. To take care of all these three requirements we introduced two types of subclusters, primary subclusters, and secondary subclusters. Primary subclusters is the one that you get by default when you create your first Eon cluster. The nodes in the primary subclusters are always up, that means they stay up and participate in the quorum. The nodes in the primary subcluster are responsible for processing commits, and also maintain a persistent copy, of catalog on disk. This is a subcluster that you would use to process all your ETL jobs, because the topper more also runs on the node, in the primary subcluster. If you want now at this point, have another subcluster, where you would like to run queries, and also, build this cluster up and down depending on the demand or the, depending on the workload, you would create a new subcluster. And this subcluster will be off-site secondary in nature. Now secondary subclusters have nodes that don't participate in quorums, so if these nodes are down, Vertica has no impact. These nodes are also not responsible for processing commit, though they maintain up-to-date copies of the catalog in memory. They don't store catalog on disk. And these are subclusters that you can add and remove very quickly, without impacting what is running on the other subclusters. We have customers running hundreds of nodes, subclusters with hundreds of nodes, and subclusters of size like 64 node, and they can bring this subcluster up and down, or add and remove, within few minutes. So before I go into the sizing of Eon Mode, I just want to say one more thing here. We are working very closely with some of our customers who are running Eon Mode and getting better feedback from that on a regular basis. And based on the feedback, we are making lots of improvements and fixes in every hot-fix that we put out. So if you are running Eon Mode, and want to be part of this group, I suggest that, you keep your cluster current with latest hot-fixes and work with us to give us feedback, and get the improvements that you need to be successful. So let's look at what there-- What we need, to size Eon clusters. Sizing Eon clusters is very different from sizing Enterprise Mode cluster. When you are running Enterprise Mode cluster or when you're sizing Vertica cluster running Enterprise Mode, you need to take into account the amount of data that you want to store, and the configuration of your node. Depending on which you decide, how many nodes you will need, and then start the cluster. In Eon Mode, to size a cluster, you need few things like, what should be your shard count. Now, shard count decides the maximum number of nodes that will participate in your query. And we'll talk about this little bit more in the next slide. You will decide on number of nodes that you will need within a subcluster, the instance type you will pick for running statistic subcluster, and how many subclusters you will need, and how many of them should be running all the time, and how many should be running in a dynamic mode. When it comes to shard count, you have to pick shard count up front, and you can't change it once your database is up and running. So, we... So, you need to pick shard count depending the number of nodes, are the same number of nodes that you will need to process a query. Now one thing that we want to remember here, is this is not amount of data that you have in database, but this is amount of data your queries will process. So, you may have data for six years, but if your queries process last month of data, on most of the occasions, or if your dashboards are processing up to six weeks, or ten minutes, based on whatever your needs are, you will decide or pick the number of shards, shard count and nodes, based on how much data your queries process. Looking at most of our customers, we think that 12 is a good number that should work for most of our customers. And, that means, the maximum number of nodes in a subcluster that will process queries is going to be 12. If you feel that, you need more than 12 nodes to process your query, you can pick other numbers like 24 or 48. If you pick a higher number, like 48, and you go with three nodes in your subcluster, that means node subscribes to 16 primary and 16 secondary shard subscription, which totals to 32 subscriptions per node. That will leave your catalog in a broken state. So, pick shard count appropriately, don't pick prime numbers, we suggest 12 should work for most of our customers, if you think you process more than, you know, the regular, the regular number that, or you think that your customers, you think your queries process terabytes of data, then pick a number like 24. Don't pick a prime number. Okay? We are also coming up with features in Vertica like current scaling, that will help you run more-- run queries on more than, more nodes than the number of shards that you pick. And that feature will be coming out soon. So if you have picked a smaller shard count, it's not the end of the story. Now, the next thing is, you need to pick how many nodes you need within your subclusters, to process your query. Ideal number would be node number equal to shard count, or, if you want to pick a number that is less, pick node count which is such that each of the nodes has a balanced distribution of subscriptions. When... So over here, you can have, option where you can have 12 nodes and 12 shards, or you can have two subclusters with 6 nodes and 12 shards. Depending on your workload, you can pick either of the two options. The first option, where you have 12 nodes and 12 shards, is more suitable for, more suitable for batch applications, whereas two subclusters with, with six nodes each, is more suitable for desktop type applications. Picking subclusters is, it depends on your workload, you can add remove nodes relative to isolation, or Elastic Throughput Scaling. Your subclusters can have nodes of different sizes, and you need to make sure that the nodes within the subcluster have to be homogenous. So this is my last slide before I hand over to Sumeet. And this I think is very important slide that I want you to pay attention to. When you pick instance, you are going to pick instance based on workload and query budget. I want to make it clear here that we want you to pay attention to the local disk, because you have depot on your local disk, which is going to be your query performance enhancer for all kinds of deployment, in cloud, as well as on premise. So you'd expect of what you read, or what you heard, depots still play a very important role in every Eon deployment, and they act like performance enhancers. Most of our customers choose Vertica because they love the performance we offer, and we don't want you to compromise on the performance. So pick nodes with some amount of local disk, at least two terabytes is what we suggest. i3 instances in Amazon have, you know, come up with a good local disk that is very helpful, and some of our customers are benefiting from. With that said, I want to pass it over to Sumeet. >> Sumeet: So, hi everyone, my name is Sumeet Keswani, and I'm a Product Technology Engineer at Vertica. I will be discussing the various use cases that customers deploy in Eon Mode. After that, I will go into some technical details of SQL, and then I'll blend that into the best practices, in Eon Mode. And finally, we'll go through some tips and tricks. So let's get started with the use cases. So a very basic use case that users will encounter, when they start Eon Mode the first time, is they will have two subclusters. The first subcluster will be the primary subcluster, used for ETL, like Shirang mentioned. And this subcluster will be mostly on, or always on. And there will be another subcluster used for, purely for queries. And this subcluster is the secondary subcluster and it will be on sometimes. Depending on the use case. Maybe from nine to five, or Monday to Friday, depending on what application is running on it, or what users are doing on it. So this is the most basic use case, something users get started with to get their feet wet. Now as the use of the deployment of Eon Mode with subcluster increases, the users will graduate into the second use case. And this is the next level of deployment. In this situation, they still have the primary subcluster which is used for ETL, typically a larger subcluster where there is more heavier ETL running, pretty much non-stop. Then they have the usual query subcluster which will use for queries, but they may add another one, another secondary subcluster for ad-hoc workloads. The motivation for this subcluster is to isolate the unpredictable workload from the predictable workload, so as not to impact certain isolates. So you may have ad-hoc queries, or users that are running larger queries or bad workloads that occur once in a while, from running on a secondary subcluster, on a different secondary subcluster, so as to not impact the more predictable workload running on the first subcluster. Now there is no reason why these two subclusters need to have the same instances, they can have different number of nodes, different instance types, different depot configurations. And everything can be different. Another benefit is, they can be metered differently, they can be costed differently, so that the appropriate user or tenant can be billed the cost of compute. Now as the use increases even further, this is what we see as the final state of a very advanced Eon Mode deployment here. As you see, there is the primary subcluster of course, used for ETL, very heavy ETL, and that's always on. There are numerous secondary subclusters, some for predictable applications that have a very fine-tuned workload that needs a definite performance. There are other subclusters that have different usages, some for ad-hoc queries, others for demanding tenants, there could be still more subclusters for different departments, like Finance, that need it maybe at the end of the quarter. So very, very different applications, and this is the full and final promise of Eon, where there is workload isolation, there is different metering, and each app runs in its own compute space. Okay, so let's talk about a very interesting feature in Eon Mode, which we call Hibernate and Revive. So what is Hibernate? Hibernating a Vertica database is the act of dissociating all the computers on the database, and shutting it down. At this point, you shut down all compute. You still pay for storage, because your data is in the S3 bucket, but all the compute has been shut down, and you do not pay for compute anymore. If you have reserved instances, or any other instances you can use them for different applications, and your Vertica database is shut down. So this is very similar to stop database, in Eon Mode, you're stopping all compute. The benefit of course being that you pay nothing anymore for compute. So what is Revive, then? The Revive is the opposite of Hibernate, where you now associate compute with your S3 bucket or your storage, and start up the database. There is one limitation here that you should be aware of, is that the size of the database that you have during Hibernate, you must revive it the same size. So if you have a 12-node primary subcluster when hibernating, you need to provision 12 nodes in order to revive. So one best practice comes down to this, is that you must shrink your database to the smallest size possible before you hibernate, so that you can revive it in the same size, and you don't have to spin up a ton of compute in order to revive. So basically, what this means is, when you have decided to hibernate, we ask you to remove all your secondary subclusters and shrink your primary subcluster down to the bare minimum before you hibernate it. And the benefit being, is when you do revive, you will have, you will be able to do so with the mimimum number of nodes. And of course, before you hibernate, you must cleanly shut down the database, so that all the data can be synced to S3. Finally, let's talk about backups and replication. Backups and replications are still supported in Eon Mode, we sometimes get the question, "We're in S3, and S3 has nine nines of reliability, we need a backup." Yes, we highly recommend backups, you can back-up by using the VBR script, you can back-up your database to another bucket, you can also copy the bucket and revive to a different, revive a different instance of your database. This is very useful because many times people want staging or development databases, and they need some of the data from production, and this is a nice way to get that. And it also makes sure that if you accidentally delete something you will be able to get back your data. Okay, so let's go into best practices now. I will start, let's talk about the depot first, which is the biggest performance enhancer that we see for queries. So, I want to state very clearly that reading from S3, or a remote object store like S3 is very slow, because data has to go over the network, and it's very expensive. You will pay for access cost. This is where S3 is not very cheap, is that every time you access the data, there is an ATI and access cost levied. Now the depot is a performance enhancing feature that will improve the performance of queries by keeping a local cache of the data that is most frequently used. It will also reduce the cost of accessing the data because you no longer have to go to the remote object store to get the data, since it's available on a local and permanent volume. Hence depot shaping is a very important aspect of performance tuning in an Eon database. What we ask you to do is, if you are going to use a specific table or partition frequency, you can choose to pin it, in the depot, so that if your depot is under pressure or is highly utilized, these objects that are most frequently used are kept in the depot. So therefore, depot, depot shaping is the act of setting eviction policies, instead you prevent the eviction of files that you believe you need to keep, so for example, you may keep the most recent year's data or the most recent, recent partition in the depot, and thereby all queries running on those partitions will be faster. At this time, we allow you to pin any table or partition in the depot, but it is not subcluster-based. Future versions of Vertica will allow you fine-tuning the depot based on each subcluster. So, let's now go and understand a little bit of internals of how a SQL query works in Eon Mode. And, once I explain this, we will blend into best practice and it will become much more clearer why we recommend certain things. So, since S3 is our layer of durability, where data is persistent in an Eon database. When you run an insert query, like, insert into table value one, or something similar. Data is synchronously written into S3. So, it will control returns back to the client, the copy of the data is first stored in the local depot, and then uploaded to S3. And only then do we hand the control back to the client. This ensures that if something bad were to happen, the data will be persistent. The second, the second types of SQL transactions are what we call DTLs, which are catalog operations. So for example, you create a table, or you added a column. These operations are actually working with metadata. Now, as you may know, S3 does not offer mutable storage, the storage in S3 is immutable. You can never append to a file in S3. And, the way transaction logs work is, they are append operation. So when you modify the metadata, you are actually appending to a transaction log. So this poses an interesting challenge which we resolve by appending to the transaction log locally in the catalog, and then there is a service that syncs the catalog to S3 every five minutes. So this poses an interesting challenge, right. If you were to destroy or delete an instance abruptly, you could lose the commits that happened in the last five minutes. And I'll speak to this more in the subsequent slides. Now, finally let's look at, drops or truncates in Eon. Now a drop or a truncate is really a combination of the first two things that we spoke about, when you drop a table, you are making, a drop operation, you are making a metadata change. You are telling Vertica that this table no longer exists, so we go into the transaction log, and append into the transaction log, that this table has been removed. This log of course, will be synced every five minutes to S3, like we spoke. There is also the secondary operation of deleting all the files that were associated with data in this table. Now these files are on S3. And we can go about deleting them synchronously, but that would take a lot of time. And we do not want to hold up the client for this duration. So at this point, we do not synchronously delete the files, we put the files that need to be removed in a reaper queue. And return the control back to the client. And this has the performance benefit as to the drops appear to occur really fast. This also has a cost benefit, batching deletes, in big batches, is more performant, and less costly. For example, on Amazon, you could delete 1,000 files at a time in a single cost. So if you batched your deletes, you could delete them very quickly. The disadvantage of this is if you were to terminate a Vertica customer abruptly, you could leak files in S3, because the reaper queue would not have had the chance to delete these files. Okay, so let's, let's go into best practices after speaking, after understanding some technical details. So, as I said, reading and writing to S3 is slow and costly. So, the first thing you can do is, avoid as many round trips to S3 as possible. The bigger the batches of data you load, the better. The better performance you get, per commit. The fact thing is, don't read and write from S3 if you can avoid it. A lot of our customers have intermediate data processing which they think temporarily they will transform the data before finally committing it. There is no reason to use regular tables for this kind of intermediate data. We recommend using local temporary tables, and local temporary tables have the benefit of not having to upload data to S3. Finally, there is another optimization you can make. Vertica has the concept of active partitions and inactive partitions. Active partitions are the ones where you have recently loaded data, and Vertica is lazy about merging these partitions into a single ROS container. Inactive partitions are historical partitions, like, consider last year's data, or the year before that data. Those partitions are aggressively merging into a single container. And how do we know how many partitions are active and inactive? Well that's based on the configuration parameter. If you load into an inactive partition, Vertica is very aggressive about merging these containers, so we download the entire partition, merge the records that you loaded into it, and upload it back again. This creates a lot of network traffic, and I said, accessing data is, from S3, slow and costly. So we recommend you not load into inactive partitions. You should load into the most recent or active partitions, and if you happen to load into inactive partitions, set your active partition count correctly. Okay, let's talk about the reaper queue. Depending on the velocity of your ETL, you can pile up a lot of files that need to be deleted asynchronously. If you were were to terminate a Vertica customer without allowing enough time for these files to get deleted, you could leak files in S3. Now, of course if you use local temporary tables this problem does not occur because the files were never created in S3, but if you are using regular tables, you must allow Vertica enough time to delete these files, and you can change the interval at which we delete, and how much time we allow to delete and shut down, by exiting some configuration parameters that I have mentioned here. And, yeah. Okay, so let's talk a little bit about a catalog at this point. So, the catalog is synced every five minutes onto S3 for persistence. And, the catalog truncation version is the minimum, minimal viable version of the catalog to which we can revive. So, for instance, if somebody destroyed a Vertica cluster, the entire Vertica cluster, the catalog truncation version is the mimimum viable version that you will be able to revive to. Now, in order to make sure that the catalog truncation version is up to date, you must always shut down your Vertica cluster cleanly. This allows the catalog to be synced to S3. Now here are some SQL commands that you can use to see what the catalog truncation version is on S3. For the most part, you don't have to worry about this if you're shutting down cleanly, so, this is only in cases of disaster or some event where all nodes were terminated, without... without the user's permission. And... And finally let's talk about backups, so one more time, we highly recommend you take backups, you know, S3 is designed for 99.9% availability, so there could be a, maybe an occasional down-time, making sure you have backups will help you if you accidentally drop a table. S3 will not protect you against data that was deleted by accident, so, having a backup helps you there. And why not backup, right, storage is cheap. You can replicate the entire bucket and have that as a backup, or have DR plus, you're running in a different region, which also sources a backup. So, we highly recommend that you make backups. So, so with this I would like to, end my presentation, and we're ready for any questions if you have it. Thank you very much. Thank you very much.

>> Paige: Hello, everybody, and thank you for joining us today for the Virtual Vertica BDC 2020. Today's breakout session is entitled "Tapping Vertica's Integration with TensorFlow for Advanced Machine Learning." I'm Paige Roberts, Opensource Relations Manager at Vertica, and I'll be your host for this session. Joining me is Vertica Software Engineer, George Larionov. >> George: Hi. >> Paige: (chuckles) That's George. So, before we begin, I encourage you guys to submit questions or comments during the virtual session. You don't have to wait. Just type your question or comment in the question box below the slides and click submit. So, as soon as a question occurs to you, go ahead and type it in, and there will be a Q and A session at the end of the presentation. We'll answer as many questions as we're able to get to during that time. Any questions we don't get to, we'll do our best to answer offline. Now, alternatively, you can visit Vertica Forum to post your questions there, after the session. Our engineering team is planning to join the forums to keep the conversation going, so you can ask an engineer afterwards, just as if it were a regular conference in person. Also, reminder, you can maximize your screen by clicking the double-arrow button in the lower right corner of the slides. And, before you ask, yes, this virtual session is being recorded, and it will be available to view by the end this week. We'll send you a notification as soon as it's ready. Now, let's get started, over to you, George. >> George: Thank you, Paige. So, I've been introduced. I'm a Software Engineer at Vertica, and today I'm going to be talking about a new feature, Vertica's Integration with TensorFlow. So, first, I'm going to go over what is TensorFlow and what are neural networks. Then, I'm going to talk about why integrating with TensorFlow is a useful feature, and, finally, I am going to talk about the integration itself and give an example. So, as we get started here, what is TensorFlow? TensorFlow is an opensource machine learning library, developed by Google, and it's actually one of many such libraries. And, the whole point of libraries like TensorFlow is to simplify the whole process of working with neural networks, such as creating, training, and using them, so that it's available to everyone, as opposed to just a small subset of researchers. So, neural networks are computing systems that allow us to solve various tasks. Traditionally, computing algorithms were designed completely from the ground up by engineers like me, and we had to manually sift through the data and decide which parts are important for the task and which are not. Neural networks aim to solve this problem, a little bit, by sifting through the data themselves, automatically and finding traits and features which correlate to the right results. So, you can think of it as neural networks learning to solve a specific task by looking through the data without having human beings have to sit and sift through the data themselves. So, there's a couple necessary parts to getting a trained neural model, which is the final goal. By the way, a neural model is the same as a neural network. Those are synonymous. So, first, you need this light blue circle, an untrained neural model, which is pretty easy to get in TensorFlow, and, in edition to that, you need your training data. Now, this involves both training inputs and training labels, and I'll talk about exactly what those two things are on the next slide. But, basically, you need to train your model with the training data, and, once it is trained, you can use your trained model to predict on just the purple circle, so new training inputs. And, it will predict the training labels for you. You don't have to label it anymore. So, a neural network can be thought of as... Training a neural network can be thought of as teaching a person how to do something. For example, if I want to learn to speak a new language, let's say French, I would probably hire some sort of tutor to help me with that task, and I would need a lot of practice constructing and saying sentences in French. And a lot of feedback from my tutor on whether my pronunciation or grammar, et cetera, is correct. And, so, that would take me some time, but, finally, hopefully, I would be able to learn the language and speak it without any sort of feedback, getting it right. So, in a very similar manner, a neural network needs to practice on, example, training data, first, and, along with that data, it needs labeled data. In this case, the labeled data is kind of analogous to the tutor. It is the correct answers, so that the network can learn what those look like. But, ultimately, the goal is to predict on unlabeled data which is analogous to me knowing how to speak French. So, I went over most of the bullets. A neural network needs a lot of practice. To do that, it needs a lot of good labeled data, and, finally, since a neural network needs to iterate over the training data many, many times, it needs a powerful machine which can do that in a reasonable amount of time. So, here's a quick checklist on what you need if you have a specific task that you want to solve with a neural network. So, the first thing you need is a powerful machine for training. We discussed why this is important. Then, you need TensorFlow installed on the machine, of course, and you need a dataset and labels for your dataset. Now, this dataset can be hundreds of examples, thousands, sometimes even millions. I won't go into that because the dataset size really depends on the task at hand, but if you have these four things, you can train a good neural network that will predict whatever result you want it to predict at the end. So, we've talked about neural networks and TensorFlow, but the question is if we already have a lot of built-in machine-learning algorithms in Vertica, then why do we need to use TensorFlow? And, to answer that question, let's look at this dataset. So, this is a pretty simple toy dataset with 20,000 points, but it shows, it simulates a more complex dataset with some sort of two different classes which are not related in a simple way. So, the existing machine-learning algorithms that Vertica already has, mostly fail on this pretty simple dataset. Linear models can't really draw a good line separating the two types of points. Naïve Bayes, also, performs pretty badly, and even the Random Forest algorithm, which is a pretty powerful algorithm, with 300 trees gets only 80% accuracy. However, a neural network with only two hidden layers gets 99% accuracy in about ten minutes of training. So, I hope that's a pretty compelling reason to use neural networks, at least sometimes. So, as an aside, there are plenty of tasks that do fit the existing machine-learning algorithms in Vertica. That's why they're there, and if one of your tasks that you want to solve fits one of the existing algorithms, well, then I would recommend using that algorithm, not TensorFlow, because, while neural networks have their place and are very powerful, it's often easier to use an existing algorithm, if possible. Okay, so, now that we've talked about why neural networks are needed, let's talk about integrating them with Vertica. So, neural networks are best trained using GPUs, which are Graphics Processing Units, and it's, basically, just a different processing unit than a CPU. GPUs are good for training neural networks because they excel at doing many, many simple operations at the same time, which is needed for a neural network to be able to iterate through the training data many times. However, Vertica runs on CPUs and cannot run on GPUs at all because that's not how it was designed. So, to train our neural networks, we have to go outside of Vertica, and exporting a small batch of training data is pretty simple. So, that's not really a problem, but, given this information, why do we even need Vertica? If we train outside, then why not do everything outside of Vertica? So, to answer that question, here is a slide that Philips was nice enough to let us use. This is an example of production system at Philips. So, it consists of two branches. On the left, we have a branch with historical device log data, and this can kind of be thought of as a bunch of training data. And, all that data goes through some data integration, data analysis. Basically, this is where you train your models, whether or not they are neural networks, but, for the purpose of this talk, this is where you would train your neural network. And, on the right, we have a branch which has live device log data coming in from various MRI machines, CAT scan machines, et cetera, and this is a ton of data. So, these machines are constantly running. They're constantly on, and there's a bunch of them. So, data just keeps streaming in, and, so, we don't want this data to have to take any unnecessary detours because that would greatly slow down the whole system. So, this data in the right branch goes through an already trained predictive model, which need to be pretty fast, and, finally, it allows Philips to do some maintenance on these machines before they actually break, which helps Philips, obviously, and definitely the medical industry as well. So, I hope this slide helped explain the complexity of a live production system and why it might not be reasonable to train your neural networks directly in the system with the live device log data. So, a quick summary on just the neural networks section. So, neural networks are powerful, but they need a lot of processing power to train which can't really be done well in a production pipeline. However, they are cheap and fast to predict with. Prediction with a neural network does not require GPU anymore. And, they can be very useful in production, so we do want them there. We just don't want to train them there. So, the question is, now, how do we get neural networks into production? So, we have, basically, two options. The first option is to take the data and export it to our machine with TensorFlow, our powerful GPU machine, or we can take our TensorFlow model and put it where the data is. In this case, let's say that that is Vertica. So, I'm going to go through some pros and cons of these two approaches. The first one is bringing the data to the analytics. The pros of this approach are that TensorFlow is already installed, running on this GPU machine, and we don't have to move the model at all. The cons, however, are that we have to transfer all the data to this machine and if that data is big, if it's, I don't know, gigabytes, terabytes, et cetera, then that becomes a huge bottleneck because you can only transfer in small quantities. Because GPU machines tend to not be that big. Furthermore, TensorFlow prediction doesn't actually need a GPU. So, you would end up paying for an expensive GPU for no reason. It's not parallelized because you just have one GPU machine. You can't put your production system on this GPU, as we discussed. And, so, you're left with good results, but not fast and not where you need them. So, now, let's look at the second option. So, the second option is bringing the analytics to the data. So, the pros of this approach are that we can integrate with our production system. It's low impact because prediction is not processor intensive. It's cheap, or, at least, it's pretty much as cheap as your system was before. It's parallelized because Vertica was always parallelized, which we'll talk about in the next slide. There's no extra data movement. You get the benefit from model management in Vertica, meaning, if you import multiple TensorFlow models, you can keep track of their various attributes, when they were imported, et cetera. And, the results are right where you need them, inside your production pipeline. So, two cons are that TensorFlow is limited to just prediction inside Vertica, and, if you want to retrain your model, you need to do that outside of Vertica and, then, reimport. So, just as a recap of parallelization. Everything in Vertica is parallelized and distributed, and TensorFlow is no exception. So, when you import your TensorFlow model to your Vertica cluster, it gets copied to all the nodes, automatically, and TensorFlow will run in fenced mode which means that it the TensorFlow process fails for whatever reason, even though it shouldn't, but if it does, Vertica itself will not crash, which is obviously important. And, finally, prediction happens on each node. There are multiple threads of TensorFlow processes running, processing different little bits of data, which is faster, much faster, than processing the data line by line because it happens all in a parallelized fashion. And, so, the result is fast prediction. So, here's an example which I hope is a little closer to what everyone is used to than the usual machine learning TensorFlow example. This is the Boston housing dataset, or, rather, a small subset of it. Now, on the left, we have the input data to go back to, I think, the first slide, and, on the right, is the training label. So, the input data consists of, each line is a plot of land in Boston, along with various attributes, such as the level of crime in that area, how much industry is in that area, whether it's on the Charles River, et cetera, and, on the right, we have as the labels the median house value in that plot of land. And, so, the goal is to put all this data into the neural network and, finally, get a model which can train... I don't know, which can predict on new incoming data and predict a good housing value for that data. Now, I'm going to go through, step by step, how to actually use TensorFlow models in Vertica. So, the first step I won't go into much detail on because there are countless tutorials and resources online on how to use TensorFlow to train a neural network, so that's the first step. Second step is to save the model in TensorFlow's 'frozen graph' format. Again, this information is available online. The third step is to create a small, simple JSON file describing the inputs and outputs of the model, and what data type they are, et cetera. And, this is needed for Vertica to be able to translate from TensorFlow land into Vertica equal land, so that it can use a sequel table instead of the input set TensorFlow usually takes. So, once you have your model file and your JSON file, you want to put both of those files in a directory on a node, any node, in a Vertica cluster, and name that directory whatever you want your model to ultimately be called inside of Vertica. So, once you do that you can go ahead and import that directory into Vertica. So, this import model's function already exists in Vertica. All we added was a new category to be able to import. So, what you need to do is specify the pass to your neural network directory and specify that the category that the model is is a TensorFlow model. Once you successfully import, in order to predict, you run this brand new predict TensorFlow function, so, in this case, we're predicting on everything from the input table, which is what the star means. The model name is Boston housing net which is the name of your directory, and, then, there's a little bit of boilerplate. And, the two ID and value after the as are just the names of the columns of your outputs, and, finally, the Boston housing data is whatever sequel table you want to predict on that fits the import type of your network. And, this will output a bunch of predictions. In this case, values of houses that the network thinks are appropriate for all the input data. So, just a quick summary. So, we talked about what is TensorFlow and what are neural networks, and, then, we discussed that TensorFlow works best on GPUs because it needs very specific characteristics. That is TensorFlow works best for training on GPUs while Vertica is designed to use CPUs, and it's really good at storing and accessing a lot of data quickly. But, it's not very well designed for having neural networks trained inside of it. Then, we talked about how neural models are powerful, and we want to use them in our production flow. And, since prediction is fast, we can go ahead and do that, but we just don't want to train there, and, finally, I presented Vertica TensorFlow integration which allows importing a trained neural model, a trained neural TensorFlow model, into Vertica and predicting on all the data that is inside Vertica with few simple lines of sequel. So, thank you for listening. I'm going to take some questions, now.