>>From the Cube Studios in Palo Alto in Boston, bringing you data-driven insights from the cube and E T R. This is breaking analysis with Dave Valante. >>Making technology predictions in 2022 was tricky business, especially if you were projecting the performance of markets or identifying I P O prospects and making binary forecast on data AI and the macro spending climate and other related topics in enterprise tech 2022, of course was characterized by a seesaw economy where central banks were restructuring their balance sheets. The war on Ukraine fueled inflation supply chains were a mess. And the unintended consequences of of forced march to digital and the acceleration still being sorted out. Hello and welcome to this week's weekly on Cube Insights powered by E T R. In this breaking analysis, we continue our annual tradition of transparently grading last year's enterprise tech predictions. And you may or may not agree with our self grading system, but look, we're gonna give you the data and you can draw your own conclusions and tell you what, tell us what you think. >>All right, let's get right to it. So our first prediction was tech spending increases by 8% in 2022. And as we exited 2021 CIOs, they were optimistic about their digital transformation plans. You know, they rushed to make changes to their business and were eager to sharpen their focus and continue to iterate on their digital business models and plug the holes that they, the, in the learnings that they had. And so we predicted that 8% rise in enterprise tech spending, which looked pretty good until Ukraine and the Fed decided that, you know, had to rush and make up for lost time. We kind of nailed the momentum in the energy sector, but we can't give ourselves too much credit for that layup. And as of October, Gartner had it spending growing at just over 5%. I think it was 5.1%. So we're gonna take a C plus on this one and, and move on. >>Our next prediction was basically kind of a slow ground ball. The second base, if I have to be honest, but we felt it was important to highlight that security would remain front and center as the number one priority for organizations in 2022. As is our tradition, you know, we try to up the degree of difficulty by specifically identifying companies that are gonna benefit from these trends. So we highlighted some possible I P O candidates, which of course didn't pan out. S NQ was on our radar. The company had just had to do another raise and they recently took a valuation hit and it was a down round. They raised 196 million. So good chunk of cash, but, but not the i p O that we had predicted Aqua Securities focus on containers and cloud native. That was a trendy call and we thought maybe an M SS P or multiple managed security service providers like Arctic Wolf would I p o, but no way that was happening in the crummy market. >>Nonetheless, we think these types of companies, they're still faring well as the talent shortage in security remains really acute, particularly in the sort of mid-size and small businesses that often don't have a sock Lacework laid off 20% of its workforce in 2022. And CO C e o Dave Hatfield left the company. So that I p o didn't, didn't happen. It was probably too early for Lacework. Anyway, meanwhile you got Netscope, which we've cited as strong in the E T R data as particularly in the emerging technology survey. And then, you know, I lumia holding its own, you know, we never liked that 7 billion price tag that Okta paid for auth zero, but we loved the TAM expansion strategy to target developers beyond sort of Okta's enterprise strength. But we gotta take some points off of the failure thus far of, of Okta to really nail the integration and the go to market model with azero and build, you know, bring that into the, the, the core Okta. >>So the focus on endpoint security that was a winner in 2022 is CrowdStrike led that charge with others holding their own, not the least of which was Palo Alto Networks as it continued to expand beyond its core network security and firewall business, you know, through acquisition. So overall we're gonna give ourselves an A minus for this relatively easy call, but again, we had some specifics associated with it to make it a little tougher. And of course we're watching ve very closely this this coming year in 2023. The vendor consolidation trend. You know, according to a recent Palo Alto network survey with 1300 SecOps pros on average organizations have more than 30 tools to manage security tools. So this is a logical way to optimize cost consolidating vendors and consolidating redundant vendors. The E T R data shows that's clearly a trend that's on the upswing. >>Now moving on, a big theme of 2020 and 2021 of course was remote work and hybrid work and new ways to work and return to work. So we predicted in 2022 that hybrid work models would become the dominant protocol, which clearly is the case. We predicted that about 33% of the workforce would come back to the office in 2022 in September. The E T R data showed that figure was at 29%, but organizations expected that 32% would be in the office, you know, pretty much full-time by year end. That hasn't quite happened, but we were pretty close with the projection, so we're gonna take an A minus on this one. Now, supply chain disruption was another big theme that we felt would carry through 2022. And sure that sounds like another easy one, but as is our tradition, again we try to put some binary metrics around our predictions to put some meat in the bone, so to speak, and and allow us than you to say, okay, did it come true or not? >>So we had some data that we presented last year and supply chain issues impacting hardware spend. We said at the time, you can see this on the left hand side of this chart, the PC laptop demand would remain above pre covid levels, which would reverse a decade of year on year declines, which I think started in around 2011, 2012. Now, while demand is down this year pretty substantially relative to 2021, I D C has worldwide unit shipments for PCs at just over 300 million for 22. If you go back to 2019 and you're looking at around let's say 260 million units shipped globally, you know, roughly, so, you know, pretty good call there. Definitely much higher than pre covid levels. But so what you might be asking why the B, well, we projected that 30% of customers would replace security appliances with cloud-based services and that more than a third would replace their internal data center server and storage hardware with cloud services like 30 and 40% respectively. >>And we don't have explicit survey data on exactly these metrics, but anecdotally we see this happening in earnest. And we do have some data that we're showing here on cloud adoption from ET R'S October survey where the midpoint of workloads running in the cloud is around 34% and forecast, as you can see, to grow steadily over the next three years. So this, well look, this is not, we understand it's not a one-to-one correlation with our prediction, but it's a pretty good bet that we were right, but we gotta take some points off, we think for the lack of unequivocal proof. Cause again, we always strive to make our predictions in ways that can be measured as accurate or not. Is it binary? Did it happen, did it not? Kind of like an O K R and you know, we strive to provide data as proof and in this case it's a bit fuzzy. >>We have to admit that although we're pretty comfortable that the prediction was accurate. And look, when you make an hard forecast, sometimes you gotta pay the price. All right, next, we said in 2022 that the big four cloud players would generate 167 billion in IS and PaaS revenue combining for 38% market growth. And our current forecasts are shown here with a comparison to our January, 2022 figures. So coming into this year now where we are today, so currently we expect 162 billion in total revenue and a 33% growth rate. Still very healthy, but not on our mark. So we think a w s is gonna miss our predictions by about a billion dollars, not, you know, not bad for an 80 billion company. So they're not gonna hit that expectation though of getting really close to a hundred billion run rate. We thought they'd exit the year, you know, closer to, you know, 25 billion a quarter and we don't think they're gonna get there. >>Look, we pretty much nailed Azure even though our prediction W was was correct about g Google Cloud platform surpassing Alibaba, Alibaba, we way overestimated the performance of both of those companies. So we're gonna give ourselves a C plus here and we think, yeah, you might think it's a little bit harsh, we could argue for a B minus to the professor, but the misses on GCP and Alibaba we think warrant a a self penalty on this one. All right, let's move on to our prediction about Supercloud. We said it becomes a thing in 2022 and we think by many accounts it has, despite the naysayers, we're seeing clear evidence that the concept of a layer of value add that sits above and across clouds is taking shape. And on this slide we showed just some of the pickup in the industry. I mean one of the most interesting is CloudFlare, the biggest supercloud antagonist. >>Charles Fitzgerald even predicted that no vendor would ever use the term in their marketing. And that would be proof if that happened that Supercloud was a thing and he said it would never happen. Well CloudFlare has, and they launched their version of Supercloud at their developer week. Chris Miller of the register put out a Supercloud block diagram, something else that Charles Fitzgerald was, it was was pushing us for, which is rightly so, it was a good call on his part. And Chris Miller actually came up with one that's pretty good at David Linthicum also has produced a a a A block diagram, kind of similar, David uses the term metacloud and he uses the term supercloud kind of interchangeably to describe that trend. And so we we're aligned on that front. Brian Gracely has covered the concept on the popular cloud podcast. Berkeley launched the Sky computing initiative. >>You read through that white paper and many of the concepts highlighted in the Supercloud 3.0 community developed definition align with that. Walmart launched a platform with many of the supercloud salient attributes. So did Goldman Sachs, so did Capital One, so did nasdaq. So you know, sorry you can hate the term, but very clearly the evidence is gathering for the super cloud storm. We're gonna take an a plus on this one. Sorry, haters. Alright, let's talk about data mesh in our 21 predictions posts. We said that in the 2020s, 75% of large organizations are gonna re-architect their big data platforms. So kind of a decade long prediction. We don't like to do that always, but sometimes it's warranted. And because it was a longer term prediction, we, at the time in, in coming into 22 when we were evaluating our 21 predictions, we took a grade of incomplete because the sort of decade long or majority of the decade better part of the decade prediction. >>So last year, earlier this year, we said our number seven prediction was data mesh gains momentum in 22. But it's largely confined and narrow data problems with limited scope as you can see here with some of the key bullets. So there's a lot of discussion in the data community about data mesh and while there are an increasing number of examples, JP Morgan Chase, Intuit, H S P C, HelloFresh, and others that are completely rearchitecting parts of their data platform completely rearchitecting entire data platforms is non-trivial. There are organizational challenges, there're data, data ownership, debates, technical considerations, and in particular two of the four fundamental data mesh principles that the, the need for a self-service infrastructure and federated computational governance are challenging. Look, democratizing data and facilitating data sharing creates conflicts with regulatory requirements around data privacy. As such many organizations are being really selective with their data mesh implementations and hence our prediction of narrowing the scope of data mesh initiatives. >>I think that was right on J P M C is a good example of this, where you got a single group within a, within a division narrowly implementing the data mesh architecture. They're using a w s, they're using data lakes, they're using Amazon Glue, creating a catalog and a variety of other techniques to meet their objectives. They kind of automating data quality and it was pretty well thought out and interesting approach and I think it's gonna be made easier by some of the announcements that Amazon made at the recent, you know, reinvent, particularly trying to eliminate ET t l, better connections between Aurora and Redshift and, and, and better data sharing the data clean room. So a lot of that is gonna help. Of course, snowflake has been on this for a while now. Many other companies are facing, you know, limitations as we said here and this slide with their Hadoop data platforms. They need to do new, some new thinking around that to scale. HelloFresh is a really good example of this. Look, the bottom line is that organizations want to get more value from data and having a centralized, highly specialized teams that own the data problem, it's been a barrier and a blocker to success. The data mesh starts with organizational considerations as described in great detail by Ash Nair of Warner Brothers. So take a listen to this clip. >>Yeah, so when people think of Warner Brothers, you always think of like the movie studio, but we're more than that, right? I mean, you think of H B O, you think of t n t, you think of C N N. We have 30 plus brands in our portfolio and each have their own needs. So the, the idea of a data mesh really helps us because what we can do is we can federate access across the company so that, you know, CNN can work at their own pace. You know, when there's election season, they can ingest their own data and they don't have to, you know, bump up against, as an example, HBO if Game of Thrones is going on. >>So it's often the case that data mesh is in the eyes of the implementer. And while a company's implementation may not strictly adhere to Jamma Dani's vision of data mesh, and that's okay, the goal is to use data more effectively. And despite Gartner's attempts to deposition data mesh in favor of the somewhat confusing or frankly far more confusing data fabric concept that they stole from NetApp data mesh is taking hold in organizations globally today. So we're gonna take a B on this one. The prediction is shaping up the way we envision, but as we previously reported, it's gonna take some time. The better part of a decade in our view, new standards have to emerge to make this vision become reality and they'll come in the form of both open and de facto approaches. Okay, our eighth prediction last year focused on the face off between Snowflake and Databricks. >>And we realized this popular topic, and maybe one that's getting a little overplayed, but these are two companies that initially, you know, looked like they were shaping up as partners and they, by the way, they are still partnering in the field. But you go back a couple years ago, the idea of using an AW w s infrastructure, Databricks machine intelligence and applying that on top of Snowflake as a facile data warehouse, still very viable. But both of these companies, they have much larger ambitions. They got big total available markets to chase and large valuations that they have to justify. So what's happening is, as we've previously reported, each of these companies is moving toward the other firm's core domain and they're building out an ecosystem that'll be critical for their future. So as part of that effort, we said each is gonna become aggressive investors and maybe start doing some m and a and they have in various companies. >>And on this chart that we produced last year, we studied some of the companies that were targets and we've added some recent investments of both Snowflake and Databricks. As you can see, they've both, for example, invested in elation snowflake's, put money into Lacework, the Secur security firm, ThoughtSpot, which is trying to democratize data with ai. Collibra is a governance platform and you can see Databricks investments in data transformation with D B T labs, Matillion doing simplified business intelligence hunters. So that's, you know, they're security investment and so forth. So other than our thought that we'd see Databricks I p o last year, this prediction been pretty spot on. So we'll give ourselves an A on that one. Now observability has been a hot topic and we've been covering it for a while with our friends at E T R, particularly Eric Bradley. Our number nine prediction last year was basically that if you're not cloud native and observability, you are gonna be in big trouble. >>So everything guys gotta go cloud native. And that's clearly been the case. Splunk, the big player in the space has been transitioning to the cloud, hasn't always been pretty, as we reported, Datadog real momentum, the elk stack, that's open source model. You got new entrants that we've cited before, like observe, honeycomb, chaos search and others that we've, we've reported on, they're all born in the cloud. So we're gonna take another a on this one, admittedly, yeah, it's a re reasonably easy call, but you gotta have a few of those in the mix. Okay, our last prediction, our number 10 was around events. Something the cube knows a little bit about. We said that a new category of events would emerge as hybrid and that for the most part is happened. So that's gonna be the mainstay is what we said. That pure play virtual events are gonna give way to hi hybrid. >>And the narrative is that virtual only events are, you know, they're good for quick hits, but lousy replacements for in-person events. And you know that said, organizations of all shapes and sizes, they learn how to create better virtual content and support remote audiences during the pandemic. So when we set at pure play is gonna give way to hybrid, we said we, we i we implied or specific or specified that the physical event that v i p experience is going defined. That overall experience and those v i p events would create a little fomo, fear of, of missing out in a virtual component would overlay that serves an audience 10 x the size of the physical. We saw that really two really good examples. Red Hat Summit in Boston, small event, couple thousand people served tens of thousands, you know, online. Second was Google Cloud next v i p event in, in New York City. >>Everything else was, was, was, was virtual. You know, even examples of our prediction of metaverse like immersion have popped up and, and and, and you know, other companies are doing roadshow as we predicted like a lot of companies are doing it. You're seeing that as a major trend where organizations are going with their sales teams out into the regions and doing a little belly to belly action as opposed to the big giant event. That's a definitely a, a trend that we're seeing. So in reviewing this prediction, the grade we gave ourselves is, you know, maybe a bit unfair, it should be, you could argue for a higher grade, but the, but the organization still haven't figured it out. They have hybrid experiences but they generally do a really poor job of leveraging the afterglow and of event of an event. It still tends to be one and done, let's move on to the next event or the next city. >>Let the sales team pick up the pieces if they were paying attention. So because of that, we're only taking a B plus on this one. Okay, so that's the review of last year's predictions. You know, overall if you average out our grade on the 10 predictions that come out to a b plus, I dunno why we can't seem to get that elusive a, but we're gonna keep trying our friends at E T R and we are starting to look at the data for 2023 from the surveys and all the work that we've done on the cube and our, our analysis and we're gonna put together our predictions. We've had literally hundreds of inbounds from PR pros pitching us. We've got this huge thick folder that we've started to review with our yellow highlighter. And our plan is to review it this month, take a look at all the data, get some ideas from the inbounds and then the e t R of January surveys in the field. >>It's probably got a little over a thousand responses right now. You know, they'll get up to, you know, 1400 or so. And once we've digested all that, we're gonna go back and publish our predictions for 2023 sometime in January. So stay tuned for that. All right, we're gonna leave it there for today. You wanna thank Alex Myerson who's on production and he manages the podcast, Ken Schiffman as well out of our, our Boston studio. I gotta really heartfelt thank you to Kristen Martin and Cheryl Knight and their team. They helped get the word out on social and in our newsletters. Rob Ho is our editor in chief over at Silicon Angle who does some great editing for us. Thank you all. Remember all these podcasts are available or all these episodes are available is podcasts. Wherever you listen, just all you do Search Breaking analysis podcast, really getting some great traction there. Appreciate you guys subscribing. I published each week on wikibon.com, silicon angle.com or you can email me directly at david dot valante silicon angle.com or dm me Dante, or you can comment on my LinkedIn post. And please check out ETR AI for the very best survey data in the enterprise tech business. Some awesome stuff in there. This is Dante for the Cube Insights powered by etr. Thanks for watching and we'll see you next time on breaking analysis.

>>from around the globe. It's the Cube presenting Cuban cloud brought to you by silicon angle. Okay, >>we're now going to explore the vision of the future of cloud computing From the perspective of one of the leaders in the field, J G >>Share >>a pure off is the vice president of As Your Data ai and Edge at Microsoft G. Welcome to the Cuban cloud. Thanks so much for participating. >>Well, thank you, Dave, and it's a real pleasure to be here with you. And I just wanna welcome the audience as well. >>Well, jg judging from your title, we have a lot of ground to cover, and our audience is definitely interested in all the topics that are implied there. So let's get right into it. You know, we've said many times in the Cube that the new innovation cocktail comprises machine intelligence or a I applied to troves of data. With the scale of the cloud. It's it's no longer, you know, we're driven by Moore's law. It's really those three factors, and those ingredients are gonna power the next wave of value creation and the economy. So, first, do you buy into that premise? >>Yes, absolutely. we do buy into it. And I think, you know, one of the reasons why we put Data Analytics and Ai together is because all of that really begins with the collection of data and managing it and governing it, unlocking analytics in it. And we tend to see things like AI, the value creation that comes from a I as being on that continues off, having started off with really things like analytics and proceeding toe. You know, machine learning and the use of data. Interesting breaks. Yes. >>I'd like to get some more thoughts around a data and how you see the future data and the role of cloud and maybe how >>Microsoft, you >>know, strategy fits in there. I mean, you, your portfolio, you got you got sequel server, Azure, Azure sequel. You got arc, which is kinda azure everywhere for people that aren't familiar with that. You've got synapse. Which course that's all the integration a data warehouse, and get things ready for B I and consumption by the business and and the whole data pipeline and a lot of other services as your data bricks you got You got cosmos in their, uh, Blockchain. You've got open source services like Post Dress and my sequel. So lots of choices there. And I'm wondering, you know, how do you think about the future of Of of Cloud data platforms? It looks like your strategies, right tool for the right job? Is that fair? >>It is fair, but it's also just to step back and look at it. It's fundamentally what we see in this market today is that customer was the Sikh really a comprehensive proposition? And when I say a comprehensive proposition, it is sometimes not just about saying that. Hey, listen way No, you're a sequel server company. We absolutely trust that you have the best Azure sequel database in the cloud, but tell us more. We've got data that's sitting in her group systems. We've got data that's sitting in Post Press in things like mongo DB, right? So that open source proposition today and data and data management and database management has become front and center, so are really sort of push. There is when it comes to migration management, modernization of data to present the broadest possible choice to our customers so we can meet them where they are. However, when it comes to analytics. One of the things they asked for is give us a lot more convergence use. You know it, really, it isn't about having 50 different services. It's really about having that one comprehensive service that is converged. That's where things like synapse Fitzer, where in just land any kind of data in the leg and then use any compute engine on top of it to drive insights from it. So, fundamentally, you know, it is that flexibility that we really sort of focus on to meet our customers where they are and really not pushing our dogma and our beliefs on it. But to meet our customers according to the way they have deployed stuff like this. >>So that's great. I want to stick on this for a minute because, you know, I know when when I have guests on like yourself, do you never want to talk about you know, the competition? But that's all we ever talk about. That's all your customers ever talk about, because because the counter to that right tool for the right job and that I would say, is really kind of Amazon's approach is is that you got the single unified data platform, the mega database that does it all. And that's kind of Oracle's approach. It sounds like you wanna have your cake and eat it, too, so you you got the right tool for the right job approach. But you've got an integration layer that allows you to have that converge database. I wonder if you could add color to that and you confirm or deny what I just said. >>No, that's a That's a very fair observation, but I I say there's a nuance in what I sort of describe when it comes to data management. When it comes to APS, we have them customers with the broadest choice. Even in that, even in that perspective, we also offer convergence. So, case in point, when you think about Cosmos TV under that one sort of service, you get multiple engines, but with the same properties, right global distribution, the five nines availability. It gives customers the ability to basically choose when they have to build that new cloud native AB toe, adopt cosmos Davey and adopted in a way that it's and choose an engine that is most flexible. Tow them, however you know when it comes to say, you know, writing a sequel server, for example from organizing it you know you want. Sometimes you just want to lift and shift it into things. Like I asked In other cases, you want to completely rewrite it, so you need to have the flexibility of choice there that is presented by a legacy off What's its on premises? When it moved into things like analytics, we absolutely believe in convergence, right? So we don't believe that look, you need to have a relation of data warehouse that is separate from a loop system that is separate from, say, a B I system. That is just, you know, it's a bolt on for us. We love the proposition off, really building things that are so integrated that once you land data, once you prep it inside the lake, you can use it for analytics. You can use it for being. You can use it for machine learning. So I think you know, are sort of differentiated. Approach speaks for itself there. Well, >>that's that's interesting, because essentially, again, you're not saying it's an either or, and you're seeing a lot of that in the marketplace. You got some companies say no, it's the Data Lake and others saying No, no put in the data warehouse and that causes confusion and complexity around the data pipeline and a lot of calls. And I'd love to get your thoughts on this. Ah, lot of customers struggled to get value out of data and and specifically data product builders of frustrated that it takes too long to go from. You know, this idea of Hey, I have an idea for a data service and it could drive monetization, but to get there, you gotta go through this complex data lifecycle on pipeline and beg people to add new data sources. And do you do you feel like we have to rethink the way that we approach data architectures? >>Look, I think we do in the cloud, and I think what's happening today and I think the place where I see the most amount of rethink the most amount of push from our customers to really rethink is the area of analytics in a I. It's almost as if what worked in the past will not work going forward. Right? So when you think about analytics on in the Enterprise today, you have relational systems, you have produced systems. You've got data marts. You've got data warehouses. You've got enterprise data warehouses. You know, those large honking databases that you use, uh, to close your books with right? But when you start to modernize it, what deep you are saying is that we don't want to simply take all of that complexity that we've built over say, you know, 34 decades and simply migrated on mass exactly as they are into the cloud. What they really want is a completely different way of looking at things. And I think this is where services like synapse completely provide a differentiated proposition to our customers. What we say there is land the data in any way you see shape or form inside the lake. Once you landed inside the lake, you can essentially use a synapse studio toe. Prep it in the way that you like, use any compute engine of your choice and and operate on this data in any way that you see fit. So, case in point, if you want to hydrate relation all data warehouse, you can do so if you want to do ad hoc analytics using something like spark. You can do so if you want to invoke power. Bi I on that data or b i on that data you can do so if you want to bring in a machine learning model on this breath data you can do so, so inherently. So when customers buy into this proposition, what it solves for them and what it gives them is complete simplicity, right? One way to land the data, multiple ways to use it. And it's all eso. >>Should we think of synapse as an abstraction layer that abstracts away the complexity of the underlying technology? Is that a fair way toe? Think about it. >>Yeah, you can think of it that way. It abstracts away, Dave a couple of things. It takes away the type of data, you know, sort of the complexities related to the type of data. It takes away the complexity related to the size of data. It takes away the complexity related to creating pipelines around all these different types of data and fundamentally puts it in a place where it can be now consumed by any sort of entity inside the actual proposition. And by that token, even data breaks. You know, you can, in fact, use data breaks in in sort off an integrated way with a synapse, Right, >>Well, so that leads me to this notion of and then wonder if you buy into it s Oh, my inference is that a data warehouse or a data lake >>could >>just be a node in inside of a global data >>mesh on. >>Then it's synapses sort of managing, uh, that technology on top. Do you buy into that that global data mesh concept >>we do. And we actually do see our customers using synapse and the value proposition that it brings together in that way. Now it's not where they start. Often times when a customer comes and says, Look, I've got an enterprise data warehouse, I want to migrate it or I have a group system. I want to migrate it. But from there, the evolution is absolutely interesting to see. I give you an example. You know, one of the customers that we're very proud off his FedEx And what FedEx is doing is it's completely reimagining its's logistics system that basically the system that delivers What is it? The three million packages a day on in doing so in this covert times, with the view of basically delivering our covert vaccines. One of the ways they're doing it is basically using synapse. Synapse is essentially that analytic hub where they can get complete view into their logistic processes. Way things are moving, understand things like delays and really put all that together in a way that they can essentially get our packages and these vaccines delivered as quickly as possible. Another example, you know, is one of my favorite, uh, we see once customers buy into it, they essentially can do other things with it. So an example of this is, uh is really my favorite story is Peace Parks Initiative. It is the premier Air White Rhino Conservancy in the world. They essentially are using data that has landed in azure images in particular. So, basically, you know, use drones over the vast area that they patrol and use machine learning on this data to really figure out where is an issue and where there isn't an issue so that this part with about 200 rangers can scramble surgically versus having to read range across the last area that they cover. So What do you see here is you know, the importance is really getting your data in order. Landed consistently. Whatever the kind of data ideas build the right pipelines and then the possibilities of transformation are just endless. >>Yeah, that's very nice how you worked in some of the customer examples. I appreciate that. I wanna ask you, though, that that some people might say that putting in that layer while it clearly adds simplification and e think a great thing that they're begins over time to be be a gap, if you will, between the ability of that layer to integrate all the primitives and all the peace parts on that, that you lose some of that fine grain control and it slows you down. What would you say to that? >>Look, I think that's what we excel at, and that's what we completely sort of buy into on. It's our job to basically provide that level off integration that granularity in the way that so it's an art, absolutely admit it's an art. There are areas where people create simplicity and not a lot of you know, sort of knobs and dials and things like that. But there are areas where customers want flexibility, right? So I think just to give you an example of both of them in landing the data inconsistency in building pipelines, they want simplicity. They don't want complexity. They don't want 50 different places to do this. Just 100 to do it. When it comes to computing and reducing this data analyzing this data, they want flexibility. This is one of the reasons why we say, Hey, listen, you want to use data breaks? If you're you're buying into that proposition and you're absolutely happy with them, you can plug plug it into it. You want to use B I and no, essentially do a small data mart. You can use B I If you say that. Look, I've landed in the lake. I really only want to use em melt, bringing your animal models and party on. So that's where the flexibility comes in. So that's sort of really sort of think about it. Well, >>I like the strategy because, you know, my one of our guest, Jim Octagon, e E. I think one of the foremost thinkers on this notion of off the data mesh and her premises that that that data builders, data product and service builders air frustrated because the the big data system is generic to context. There's no context in there. But by having context in the big data architecture and system, you could get products to market much, much, much faster. So but that seems to be your philosophy. But I'm gonna jump ahead to do my ecosystem question. You've mentioned data breaks a couple of times. There's another partner that you have, which is snowflake. They're kind of trying to build out their own, uh, data cloud, if you will, on global mesh in and the one hand, their partner. On the other hand, there are competitors. How do you sort of balance and square that circle? >>Look, when I see snowflake, I actually see a partner. You know that when we essentially you know, we are. When you think about as you know, this is where I sort of step back and look at Azure as a whole and in azure as a whole. Companies like snowflakes are vital in our ecosystem, right? I mean, there are places we compete, but you know, effectively by helping them build the best snowflake service on Asia. We essentially are able toe, you know, differentiate and offer a differentiated value proposition compared to, say, a Google or on AWS. In fact, that's being our approach with data breaks as well, where you know they are effectively on multiple club, and our opportunity with data breaks is toe essentially integrate them in a way where we offer the best experience. The best integrations on Azure Barna That's always been a focus. >>That's hard to argue with. Strategy. Our data with our data partner eat er, shows Microsoft is both pervasive and impressively having a lot of momentum spending velocity within the budget cycles. I wanna come back thio ai a little bit. It's obviously one of the fastest growing areas in our in our survey data. As I said, clearly, Microsoft is a leader in this space. What's your what's your vision of the future of machine intelligence and how Microsoft will will participate in that opportunity? >>Yeah, so fundamentally, you know, we've built on decades of research around, you know, around, you know, essentially, you know, vision, speech and language that's being the three core building blocks and for the for a really focused period of time we focused on essentially ensuring human parody. So if you ever wondered what the keys to the kingdom are it, czar, it's the most we built in ensuring that the research posture that we've taken there, what we then done is essentially a couple of things we focused on, essentially looking at the spectrum. That is a I both from saying that, Hollis and you know it's gotta work for data. Analysts were looking toe basically use machine learning techniques, toe developers who are essentially, you know, coding and building a machine learning models from scratch. So for that select proposition manifesto us, as you know, really a. I focused on all skill levels. The other court thing we've done is that we've also said, Look, it will only work as long as people trust their data and they can trust their AI models. So there's a tremendous body of work and research we do in things like responsibility. So if you ask me where we sort of push on is fundamentally to make sure that we never lose sight of the fact that the spectrum off a I, and you can sort of come together for any skill level, and we keep that responsibly. I proposition. Absolutely strong now against that canvas, Dave. I'll also tell you that you know, as edge devices get way more capable, right where they can input on the edge, see a camera or a mike or something like that, you will see us pushing a lot more of that capability onto the edge as well. But to me, that's sort of a modality. But the core really is all skill levels and that responsible denia. >>Yeah, So that that brings me to this notion of wanna bring an edge and and hybrid cloud Understand how you're thinking about hybrid cloud multi cloud. Obviously one of your competitors, Amazon won't even say the word multi cloud you guys have, Ah, you know, different approach there. But what's the strategy with regard? Toe, toe hybrid. You know, Do you see the cloud you bringing azure to the edge? Maybe you could talk about that and talk about how you're different from the competition. >>Yeah, I think in the edge from Annette, you know, I live in I'll be the first one to say that the word nge itself is conflated. Okay, It's, uh but I will tell you, just focusing on hybrid. This is one of the places where you know I would say the 2020 if I would have looked back from a corporate perspective. In particular, it has Bean the most informative because we absolutely saw customers digitizing moving to the cloud. And we really saw hybrid in action. 2020 was the year that hybrid sort of really became really from a cloud computing perspective and an example of this is we understood it's not all or nothing. So sometimes customers want azure consistency in their data centers. This is where things like Azure stack comes in. Sometimes they basically come to us and say, We want the flexibility of adopting flexible pattern, you know, platforms like, say, containers orchestra, Cuban Pettis, so that we can essentially deployed wherever you want. And so when we design things like art, it was built for that flexibility in mind. So here is the beauty of what's something like our can do for you. If you have a kubernetes endpoint anywhere we can deploy and as your service onto it, that is the promise, which means if for some reason, the customer says that. Hey, I've got this kubernetes endpoint in AWS and I love as your sequel. You will be able to run as your sequel inside AWS. There's nothing that stops you from doing it so inherently you remember. Our first principle is always to meet our customers where they are. So from that perspective, multi cloud is here to stay. You know, we're never going to be the people that says, I'm sorry, we will never see a But it is a reality for our customers. >>So I wonder if we could close. Thank you for that by looking, looking back and then and then ahead. And I wanna e wanna put forth. Maybe it's, Ah criticism, but maybe not. Maybe it's an art of Microsoft, but But first you know, you get Microsoft an incredible job of transitioning. It's business as your nominee president Azzawi said. Our data shows that so two part question First, Microsoft got there by investing in the cloud, really changing its mind set, I think, in leveraging its huge software state and customer base to put Azure at the center of its strategy, and many have said me included that you got there by creating products that air Good enough. You know, we do a 1.0, it's not that great. And the two Dato, and maybe not the best, but acceptable for your customers. And that's allowed you to grow very rapidly expanding market. >>How >>do you respond to that? Is that is that a fair comment? Ume or than good enough? I wonder if you could share your >>thoughts, gave you? You hurt my feelings with that question. I don't hate me, g getting >>it out there. >>So there was. First of all, thank you for asking me that. You know, I am absolutely the biggest cheerleader. You'll find a Microsoft. I absolutely believe you know that I represent the work off almost 9000 engineers and we wake up every day worrying about our customer and worrying about the customer condition and toe. Absolutely. Make sure we deliver the best in the first time that we do. So when you take the platter off products we've delivered in nausea, be it as your sequel, be it as your cosmos TV synapse as your data breaks, which we did in partnership with data breaks, a za machine learning and recently when we prevail, we sort off, you know, sort of offered the world's first comprehensive data government solution in azure purview. I would humbly submit to you that we're leading the way and we're essentially showing how the future off data ai and the actual work in the cloud. >>I'd be disappointed if you if you had If you didn't, if you capitulated in any way J g So so thank you for that. And the kind of last question is, is looking forward and how you're thinking about the future of cloud last decade. A lot about your cloud migration simplifying infrastructure management, deployment SAS if eyeing my enterprise, lot of simplification and cost savings. And, of course, the redeployment of resource is toward digital transformation. Other other other valuable activities. How >>do >>you think this coming decade will will be defined? Will it be sort of more of the same? Or is there Is there something else out there? >>I think I think that the coming decade will be one where customers start one law outside value out of this. You know what happened in the last decade when people leave the foundation and people essentially looked at the world and said, Look, we've got to make the move, you know, the largely hybrid, but we're going to start making steps to basically digitize and modernize our platforms. I would tell you that with the amount of data that people are moving to the cloud just as an example, you're going to see use of analytics ai for business outcomes explode. You're also going to see a huge sort of focus on things like governance. You know, people need to know where the data is, what the data catalog continues, how to govern it, how to trust this data and given all other privacy and compliance regulations out there. Essentially, they're complying this posture. So I think the unlocking of outcomes versus simply Hey, I've saved money Second, really putting this comprehensive sort off, you know, governance, regime in place. And then, finally, security and trust. It's going to be more paramount than ever before. Yeah, >>nobody's gonna use the data if they don't trust it. I'm glad you brought up your security. It's It's a topic that hits number one on the CEO list. J G. Great conversation. Obviously the strategy is working, and thanks so much for participating in Cuba on cloud. >>Thank you. Thank you, David. I appreciate it and thank you to. Everybody was tuning in today. >>All right? And keep it right there. I'll be back with our next guest right after this short break.

>>from around the >>globe. It's the Cube with digital coverage of AWS reinvent 2020 sponsored by Intel and AWS. Yeah, >>Hi. We are the Cube live covering AWS reinvent 2020. I'm Lisa Martin, and I've got one of our cube alumni back with me. Mike Miller is here. General manager of A W s AI Devices at AWS. Mike, welcome back to the Cube. >>Hi, Lisa. Thank you so much for having me. It's really great to join you all again at this virtual reinvent. >>Yes, I think last year you were on set. We have always had to. That's at reinvent. And you you had the deep race, your car, and so we're obviously socially distance here. But talk to me about deepracer. What's going on? Some of the things that have gone on the last year that you're excited >>about. Yeah, I'd love to tell. Tell you a little bit about what's been happening. We've had a tremendous year. Obviously, Cove. It has restricted our ability to have our in person races. Eso we've really gone gone gangbusters with our virtual league. So we have monthly races for competitors that culminate in the championship. Um, at reinvent. So this year we've got over 100 competitors who have qualified and who are racing virtually with us this year at reinvent. They're participating in a series of knockout rounds that are being broadcast live on twitch over the next week. That will whittle the group down to AH Group of 32 which will have a Siris of single elimination brackets leading to eight finalists who will race Grand Prix style five laps, eight cars on the track at the same time and will crown the champion at the closing keynote on December 15th this year. >>Exciting? So you're bringing a reinforcement, learning together with with sports that so many of us have been missing during the pandemic. We talked to me a little bit about some of the things that air that you've improved with Deep Racer and some of the things that are coming next year. Yeah, >>absolutely so, First of all, Deep Racer not only has been interesting for individuals to participate in the league, but we continue to see great traction and adoption amongst big customers on dare, using Deep Racer for hands on learning for machine learning, and many of them are turning to Deep Racer to train their workforce in machine learning. So over 150 customers from the likes of Capital One Moody's, Accenture, DBS Bank, JPMorgan Chase, BMW and Toyota have held Deep Racer events for their workforces. And in fact, three of those customers Accenture, DBS Bank and J. P. Morgan Chase have each trained over 1000 employees in their organization because they're just super excited. And they find that deep racers away to drive that excitement and engagement across their customers. We even have Capital one expanded this to their families, so Capital One ran a deep raise. Their Kids Cup, a family friendly virtual competition this past year were over. 250 Children and 200 families got to get hands on with machine learning. >>So I envisioned some. You know, this being a big facilitator during the pandemic when there's been this massive shift to remote work has have you seen an uptick in it for companies that talking about training need to be ableto higher? Many, many more people remotely but also train them? Is deep Racer facilitator of that? Yeah, >>absolutely. Deep Racer has ah core component of the experience, which is all virtualized. So we have, ah, console and integration with other AWS services so that racers can participate using a three d racing simulator. They can actually see their car driving around a track in a three D world simulation. Um, we're also selling the physical devices. So you know, if participants want to get the one of those devices and translate what they've done in the virtual world to the real world, they can start doing that. And in fact, just this past year, we made our deep race or car available for purchase internationally through the Amazon Com website to help facilitate that. >>So how maney deep racers air out there? I'm just curious. >>Oh, thousands. Um, you know, And there what? What we've seen is some companies will purchase you, know them in bulk and use them for their internal leagues. Just like you know, JP Morgan Chase on DBS Bank. These folks have their own kind of tracks and racers that they'll use to facilitate both in person as well as the virtual racing. >>I'm curious with this shift to remote that we mentioned a minute ago. How are you seeing deepracer as a facilitator of engagement. You mentioned engagement. And that's one of the biggest challenges that so Maney teams develops. Processes have without being co located with each other deep Brister help with that. I mean, from an engagement perspective, I think >>so. What we've seen is that Deep Racer is just fun to get your hands on. And we really lower the learning curve for machine learning. And in particular, this branch called reinforcement Learning, which is where you train this agent through trial and error toe, learn how to do a new, complex task. Um, and what we've seen is that customers who have introduced Deep Racer, um, as an event for their employees have seen ah, very wide variety of employees. Skill sets, um, kind of get engaged. So you've got not just the hardcore deep data scientists or the M L engineers. You've got Web front end programmers. You even have some non technical folks who want to get their hands dirty. Onda learn about machine learning and Deep Racer really is a nice, gradual introduction to doing that. You can get engaged with it with very little kind of coding knowledge at all. >>So talk to me about some of the new services. And let's look at some specific use case customer use cases with each service. Yeah, >>absolutely. So just to set the context. You know, Amazon's got hundreds. A ws has hundreds of thousands of customers doing machine learning on AWS. No customers of all sizes are embedding machine learning into their no core business processes. And one of the things that we always do it Amazon is We're listening to customers. You know, 90 to 95% of our road maps are driven by customer feedback. And so, as we've been talking to these industrial manufacturing customers, they've been telling us, Hey, we've got data. We've got these processes that are happening in our industrial sites. Um, and we just need some help connecting the dots like, how do we really most effectively use machine learning to improve our processes in these industrial and manufacturing sites? And so we've come up with these five services. They're focused on industrial manufacturing customers, uh, two of the services air focused around, um, predictive maintenance and, uh, the other three services air focused on computer vision. Um, and so let's start with the predictive maintenance side. So we announced Amazon Monitor On and Amazon look out for equipment. So these services both enable predictive maintenance powered by machine learning in a way that doesn't require the customer to have any machine learning expertise. So Mono Tron is an end to end machine learning system with sensors, gateway and an ML service that can detect anomalies and predict when industrial equipment will require maintenance. I've actually got a couple examples here of the sensors in the gateway, so this is Amazon monitor on these little sensors. This little guy is a vibration and temperature sensor that's battery operated, and wireless connects to the gateway, which then transfers the data up to the M L Service in the cloud. And what happens is, um, the sensors can be connected to any rotating machinery like pump. Pour a fan or a compressor, and they will send data up to the machine learning cloud service, which will detect anomalies or sort of irregular kind of sensor readings and then alert via a mobile app. Just a tech or a maintenance technician at an industrial site to go have a look at their equipment and do some preventative maintenance. So um, it's super extreme line to end to end and easy for, you know, a company that has no machine learning expertise to take advantage of >>really helping them get on board quite quickly. Yeah, >>absolutely. It's simple tea set up. There's really very little configuration. It's just a matter of placing the sensors, pairing them up with the mobile app and you're off and running. >>Excellent. I like easy. So some of the other use cases? Yeah, absolutely. >>So So we've seen. So Amazon fulfillment centers actually have, um, enormous amounts of equipment you can imagine, you know, the size of an Amazon fulfillment center. 28 football fields, long miles of conveyor belts and Amazon fulfillment centers have started to use Amazon monitor on, uh, to monitor some of their conveyor belts. And we've got a filament center in Germany that has started using these 1000 sensors, and they've already been able to, you know, do predictive maintenance and prevent downtime, which is super costly, you know, for businesses, we've also got customers like Fender, you know, who makes guitars and amplifiers and musical equipment. Here in the US, they're adopting Amazon monitor on for their industrial machinery, um, to help prevent downtime, which again can cost them a great deal as they kind of hand manufacture these high end guitars. Then there's Amazon. Look out for equipment, which is one step further from Amazon monitor on Amazon. Look out for equipment. Um provides a way for customers to send their own sensor data to AWS in order to build and train a model that returns predictions for detecting abnormal equipment behavior. So here we have a customer, for example, like GP uh, E P s in South Korea, or I'm sorry, g S E P s in South Korea there in industrial conglomerate, and they've been collecting their own data. So they have their own sensors from industrial equipment for a decade. And they've been using just kind of rule basic rules based systems to try to gain insight into that data. Well, now they're using Amazon, look out for equipment to take all of their existing sensor data, have Amazon for equipment, automatically generate machine learning models on, then process the sensor data to know when they're abnormalities or when some predictive maintenance needs to occur. >>So you've got the capabilities of working with with customers and industry that that don't have any ML training to those that do have been using sensors. So really, everybody has an opportunity here to leverage this new Amazon technology, not only for predicted, but one of the things I'm hearing is contact list, being able to understand what's going on without having to have someone physically there unless there is an issue in contact. This is not one of the words of 2020 but I think it probably should be. >>Yeah, absolutely. And in fact, that that was some of the genesis of some of the next industrial services that we announced that are based on computer vision. What we saw on what we heard when talking to these customers is they have what we call human inspection processes or manual inspection processes that are required today for everything from, you know, monitoring you like workplace safety, too, you know, quality of goods coming off of a machinery line or monitoring their yard and sort of their, you know, truck entry and exit on their looking for computer vision toe automate a lot of these tasks. And so we just announced a couple new services that use computer vision to do that to automate these once previously manual inspection tasks. So let's start with a W A. W s Panorama uses computer vision toe improve those operations and workplace safety. AWS Panorama is, uh, comes in two flavors. There's an appliance, which is, ah, box like this. Um, it basically can go get installed on your network, and it will automatically discover and start processing the video feeds from existing cameras. So there's no additional capital expense to take a W s panorama and have it apply computer vision to the cameras that you've already got deployed, you know, So customers are are seeing that, um, you know, computer vision is valuable, but the reason they want to do this at the edge and put this computer vision on site is because sometimes they need to make very low Leighton see decisions where if you have, like a fast moving industrial process, you can use computer vision. But I don't really want to incur the cost of sending data to the cloud and back. I need to make a split second decision, so we need machine learning that happens on premise. Sometimes they don't want to stream high bandwidth video. Or they just don't have the bandwidth to get this video back to the cloud and sometimes their data governance or privacy restrictions that restrict the company's ability to send images or video from their site, um, off site to the cloud. And so this is why Panorama takes this machine learning and makes it happen right here on the edge for customers. So we've got customers like Cargill who uses or who is going to use Panorama to improve their yard management. They wanna use computer vision to detect the size of trucks that drive into their granaries and then automatically assign them to an appropriately sized loading dock. You've got a customer like Siemens Mobility who you know, works with municipalities on, you know, traffic on by other transport solutions. They're going to use AWS Panorama to take advantage of those existing kind of traffic cameras and build machine learning models that can, you know, improve congestion, allocate curbside space, optimize parking. We've also got retail customers. For instance, Parkland is a Canadian fuel station, um, and retailer, you know, like a little quick stop, and they want to use Panorama to do things like count the people coming in and out of their stores and do heat maps like, Where are people visiting my store so I can optimize retail promotions and product placement? >>That's fantastic. The number of use cases is just, I imagine if we had more time like you could keep going and going. But thank you so much for not only sharing what's going on with Deep Racer and the innovations, but also for show until even though we weren't in person at reinvent this year, Great to have you back on the Cube. Mike. We appreciate your time. Yeah, thanks, Lisa, for having me. I appreciate it for Mike Miller. I'm Lisa Martin. You're watching the cubes Live coverage of aws reinvent 2020.

>>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.

>>Hi everyone, This system A fellow from the University of Tokyo before I thought that would like to thank you she and 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 or some of the recent works that have been done either by me or by character of Hong Kong Noise Group indicating the title of my talk is a neuro more fic in silica simulator for the commenters 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 I will show some proof of concept of the game in performance that can be obtained using dissimulation in the second part and the production 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 adapted from a recent natural tronics paper from the Village Back 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 purification machine, or a recently proposed restricted Bozeman machine, FPD eight, 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 influx you beat or the energy efficiency off memory sisters uh P. J. O are still an attractive platform for building large theorizing 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 particle 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 f. D. A s. 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 suggesting 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 neck 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 a 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 Cortes in machine, which is a growing 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 cooking cm using oh, more than detection and refugee A then injection off the cooking time and eso this dynamics in both cases of CME 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 rising 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 convergence to the global minimum of there's even 20 and using this approach. And so this is >>why we propose toe introduce a macro structure the system or where one analog spin or one D o. P. O is replaced by a pair off one and knock spin and one error on cutting. 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 to force the amplitude of the expense toe, become equal to certain target amplitude. A Andi. This is known by moderating the strength off the icing complaints or see the the error variable e I multiply the icing complain here in the dynamics off UH, 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 think introduces a >>symmetry in the system, which in turn creates chaotic dynamics, which I'm showing here for solving certain current size off, um, escape problem, Uh, in which the exiled from here in the i r. From here and the value of the icing energy is shown in the bottom plots. And you see this Celtics search that visit various local minima of the as Newtonian and eventually finds the local minima Um, >>it can be shown that this modulation off the target opportunity can be used to destabilize all the local minima off the icing hamiltonian 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 the limits of contractors or quality contractors. They can also be destabilized using a moderation of the target amplitude. And so we have proposed in the past two different motivation of the target constitute the first one is a moderation that ensure the 100 >>reproduction rate of the system to become positive on this forbids the creation of any non tree retractors. And but in this work I will talk about another modulation or Uresti 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 the current 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 eso 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 everything. Yet for a system with 500 spins analog 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 tobacco cm 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 gear repression to replicate the post phaser 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, all the dog products, respect to the problem size. And and if we had a new infinite amount of resources and PGA to simulate the dynamics, then the non in optical 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 low carrot off end and while the kite off end. Because computing the dot product involves the summing, all the terms in the products, which is done by a nephew, Jay by another tree, which heights scares a 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 and over you and for the products in the end of you square eso typically for low NF pdf cheap P a. You know 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 started 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 Yeah, click the extra co components within the F p G A in order which is shown here in this right panel here in order to minimize the finding finance of the system and to minimize the long distance that the path in the in the fpt So I'm not going to the details of how this is implemented the PGA. But just to give you a new idea off why the Iraqi Yahiko organization off the system becomes extremely important toe get good performance for simulator organizing mission. 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 result for solving escape problems, free connected person, randomly person minus one, spin 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 Nina successful BT against the problem size here and and in red here there's propose F B J 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. It's similar to the car testing machine >>and security. You see that the scaling off the numbers of metrics victor 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 is possibility 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 center dining in purple, for example, and So you see that the scaring off this purpose simulator is is rather good and that for larger politicizes, we can get orders of magnitude faster than the state of the other approaches. >>Moreover, the relatively good scanning off the time to search in respect to problem size uh, they indicate that the FBT implementation would be faster than risk Other recently proposed izing machine, such as the Hope you know network implemented on Memory Sisters. 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 implemented a PGA proposed by some group in Brooklyn recently again, which is very fast for small promise sizes. But which canning is bad So that, uh, this worse than the purpose approach so that we can expect that for promise sizes larger than, let's say, 1000 spins. The purpose, 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 cut values 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 recount. 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 trickle 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 problems respect to the prime assize. And here, compared to different with such publicizing machines, particularly the digital and, you know, free to is shown in the green here, the green >>line without that's and, uh and we should two different, uh, prosthesis 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 showed that we probably can solve Prime Escape Program of Science 2000 spins 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 or 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 out 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 classical 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 eight f p g. A. In which we will add the more refined models truncated bigger in the bottom question model that just talked about on the supports in which he valued waits for the rising problems and support the cement. So we will announce >>later when this is available, and Farah is working hard to get the first version available sometime in September. Thank you all, and we'll be happy to answer any questions that you have.

>>Good morning. Good afternoon. Good evening, everyone. I should thank Entity 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 should 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 meta 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 G I J. 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 in 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 face 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 strength 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 face their frequency lock to the pump. But it can also lead in either the zero pie face state 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. If you have any 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. We'll 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 thin 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 oscillating 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 on our 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 is both in the linear and >>nonlinear 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 of 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. Yeah, 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, and 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 Hamiltonian model A. So the optical loss 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 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 this 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 their dynamics 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 on 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 we were 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 to 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 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 oh, classical and quantum, non innate behaviors in these networks. >>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 phase 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 Opio 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 in the non the general 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. 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 of 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 oppose. And that's a very abrupt face transition and compared to the to the single Opio face 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 linear elements, where we are now with the optics is probably very similar to seven years ago, which is a tabletop 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 Did 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 non in your process is 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 lithium 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 OPI ohs and the Opio networks are in the works, and that's not the only way of making large networks. 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 in o pos, which is can we have the quantum superposition of >>the zero pie states that I talked about >>and the nano photonics within? I would provide 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, other 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 Hamiltonian implementations on those networks. So if you can't build a pos, 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 Pippen O 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 on 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 stop here and thank you for your attention.

>>Hi, everyone. My name is my uncle. My uncle Poor I worked with Geo reminds you in India. We call ourselves Geo Platforms. Now on. We've been recently in the news. You've raised a lot off funding from one of the largest, most of the largest tech companies in the world. And I'm here to talk about Geos Cloud Journey, Onda Mantis Partnership. I've titled it the story often, Underdog becoming the largest telecom company in India within four years, which is really special. And we're, of course, held by the cloud. So quick disclaimer. Right. The content shared here is only for informational purposes. Um, it's only for this event. And if you want to share it outside, especially on social media platforms, we need permission from Geo Platforms limited. Okay, quick intro about myself. I am a VP of engineering a geo. I lead the Cloud Services and Platforms team with NGO Andi. I mean the geo since the beginning, since it started, and I've seen our cloud footprint grow from a handful of their models to now eight large application data centers across three regions in India. And we'll talk about how we went here. All right, Let's give you an introduction on Geo, right? Giorgio is on how we became the largest telecom campaign, India within four years from 0 to 400 million subscribers. And I think there are There are a lot of events that defined Geo and that will give you an understanding off. How do you things and what you did to overcome massive problems in India. So the slide that I want to talkto is this one and, uh, I The headline I've given is, It's the Geo is the fastest growing tech company in the world, which is not a new understatement. It's eggs, actually, quite literally true, because very few companies in the world have grown from zero to 400 million subscribers within four years paying subscribers. And I consider Geo Geos growth in three phases, which I have shown on top. The first phase we'll talk about is how geo grew in the smartphone market in India, right? And what we did to, um to really disrupt the telecom space in India in that market. Then we'll talk about the feature phone phase in India and how Geo grew there in the future for market in India. and then we'll talk about what we're doing now, which we call the Geo Platforms phase. Right. So Geo is a default four g lt. Network. Right. So there's no to geo three g networks that Joe has, Um it's a state of the art four g lt voiceover lt Network and because it was designed fresh right without any two D and three G um, legacy technologies, there were also a lot of challenges Lawn geo when we were starting up. One of the main challenges waas that all the smart phones being sold in India NGOs launching right in 2000 and 16. They did not have the voice or lt chip set embedded in the smartphone because the chips it's far costlier to embed in smartphones and India is a very price and central market. So none of the manufacturers were embedding the four g will teach upset in the smartphones. But geos are on Lee a volte in network, right for the all the network. So we faced a massive problem where we said, Look there no smartphones that can support geo. So how will we grow Geo? So in order to solve that problem, we launched our own brand of smartphones called the Life um, smartphones. And those phones were really high value devices. So there were $50 and for $50 you get you You At that time, you got a four g B storage space. A nice big display for inch display. Dual cameras, Andi. Most importantly, they had volte chip sets embedded in them. Right? And that got us our initial customers the initial for the launch customers when we launched. But more importantly, what that enabled other oh, EMS. What that forced the audience to do is that they also had to launch similar smartphones competing smartphones with voltage upset embedded in the same price range. Right. So within a few months, 3 to 4 months, um, all the other way EMS, all the other smartphone manufacturers, the Samsung's the Micromax is Micromax in India, they all had volte smartphones out in the market, right? And I think that was one key step We took off, launching our own brand of smartphone life that helped us to overcome this problem that no smartphone had. We'll teach upsets in India and then in order. So when when we were launching there were about 13 telecom companies in India. It was a very crowded space on demand. In order to gain a foothold in that market, we really made a few decisions. Ah, phew. Key product announcement that really disrupted this entire industry. Right? So, um, Geo is a default for GLT network itself. All I p network Internet protocol in everything. All data. It's an all data network and everything from voice to data to Internet traffic. Everything goes over this. I'll goes over Internet protocol, and the cost to carry voice on our smartphone network is very low, right? The bandwidth voice consumes is very low in the entire Lt band. Right? So what we did Waas In order to gain a foothold in the market, we made voice completely free, right? He said you will not pay anything for boys and across India, we will not charge any roaming charges across India. Right? So we made voice free completely and we offer the lowest data rates in the world. We could do that because we had the largest capacity or to carry data in India off all the other telecom operators. And these data rates were unheard off in the world, right? So when we launched, we offered a $2 per month or $3 per month plan with unlimited data, you could consume 10 gigabytes of data all day if you wanted to, and some of our subscriber day. Right? So that's the first phase off the overgrowth and smartphones and that really disorders. We hit 100 million subscribers in 170 days, which was very, very fast. And then after the smartphone faith, we found that India still has 500 million feature phones. And in order to grow in that market, we launched our own phone, the geo phone, and we made it free. Right? So if you take if you took a geo subscription and you carried you stayed with us for three years, we would make this phone tree for your refund. The initial deposit that you paid for this phone and this phone had also had quite a few innovations tailored for the Indian market. It had all of our digital services for free, which I will talk about soon. And for example, you could plug in. You could use a cable right on RCR HDMI cable plug into the geo phone and you could watch TV on your big screen TV from the geophones. You didn't need a separate cable subscription toe watch TV, right? So that really helped us grow. And Geo Phone is now the largest selling feature phone in India on it. 100 million feature phones in India now. So now now we're in what I call the geo platforms phase. We're growing of a geo fiber fiber to the home fiber toe the office, um, space. And we've also launched our new commerce initiatives over e commerce initiatives and were steadily building platforms that other companies can leverage other companies can use in the Jeon o'clock. Right? So this is how a small startup not a small start, but a start of nonetheless least 400 million subscribers within four years the fastest growing tech company in the world. Next, Geo also helped a systemic change in India, and this is massive. A lot of startups are building on this India stack, as people call it, and I consider this India stack has made up off three things, and the acronym I use is jam. Trinity, right. So, um, in India, systemic change happened recently because the Indian government made bank accounts free for all one billion Indians. There were no service charges to store money in bank accounts. This is called the Jonathan. The J. GenDyn Bank accounts. The J out off the jam, then India is one of the few countries in the world toe have a digital biometric identity, which can be used to verify anyone online, which is huge. So you can simply go online and say, I am my ankle poor on duh. I verify that this is indeed me who's doing this transaction. This is the A in the jam and the last M stands for Mobil's, which which were held by Geo Mobile Internet in a plus. It is also it is. It also stands for something called the U. P I. The United Unified Payments Interface. This was launched by the Indian government, where you can carry digital transactions for free. You can transfer money from one person to the to another, essentially for free for no fee, right so I can transfer one group, even Indian rupee to my friend without paying any charges. That is huge, right? So you have a country now, which, with a with a billion people who are bank accounts, money in the bank, who you can verify online, right and who can pay online without any problems through their mobile connections held by G right. So suddenly our market, our Internet market, exploded from a few million users to now 506 106 100 million mobile Internet users. So that that I think, was a massive such a systemic change that happened in India. There are some really large hail, um, numbers for this India stack, right? In one month. There were 1.6 billion nuclear transactions in the last month, which is phenomenal. So next What is the impact of geo in India before you started, we were 155th in the world in terms off mobile in terms of broadband data consumption. Right. But after geo, India went from one 55th to the first in the world in terms of broadband data, largely consumed on mobile devices were a mobile first country, right? We have a habit off skipping technology generation, so we skip fixed line broadband and basically consuming Internet on our mobile phones. On average, Geo subscribers consumed 12 gigabytes of data per month, which is one of the highest rates in the world. So Geo has a huge role to play in making India the number one country in terms off broad banded consumption and geo responsible for quite a few industry first in the telecom space and in fact, in the India space, I would say so before Geo. To get a SIM card, you had to fill a form off the physical paper form. It used to go toe Ah, local distributor. And that local distributor is to check the farm that you feel incorrectly for your SIM card and then that used to go to the head office and everything took about 48 hours or so, um, to get your SIM card. And sometimes there were problems there also with a hard biometric authentication. We enable something, uh, India enable something called E K Y C Elektronik. Know your customer? We took a fingerprint scan at our point of Sale Reliance Digital stores, and within 15 minutes we could verify within a few minutes. Within a few seconds we could verify that person is indeed my hunk, right, buying the same car, Elektronik Lee on we activated the SIM card in 15 minutes. That was a massive deal for our growth. Initially right toe onboard 100 million customers. Within our and 70 days. We couldn't have done it without be K. I see that was a massive deal for us and that is huge for any company starting a business or start up in India. We also made voice free, no roaming charges and the lowest data rates in the world. Plus, we gave a full suite of cloud services for free toe all geo customers. For example, we give goTV essentially for free. We give GOTV it'll law for free, which people, when we have a launching, told us that no one would see no one would use because the Indians like watching TV in the living rooms, um, with the family on a big screen television. But when we actually launched, they found that GOTV is one off our most used app. It's like 70,000,080 million monthly active users, and now we've basically been changing culture in India where culture is on demand. You can watch TV on the goal and you can pause it and you can resume whenever you have some free time. So really changed culture in India, India on we help people liver, digital life online. Right, So that was massive. So >>I'm now I'd like to talk about our cloud >>journey on board Animal Minorities Partnership. We've been partners that since 2014 since the beginning. So Geo has been using open stack since 2014 when we started with 14 note luster. I'll be one production environment One right? And that was I call it the first wave off our cloud where we're just understanding open stack, understanding the capabilities, understanding what it could do. Now we're in our second wave. Where were about 4000 bare metal servers in our open stack cloud multiple regions, Um, on that around 100,000 CPU cores, right. So it's a which is one of the bigger clouds in the world, I would say on almost all teams, with Ngor leveraging the cloud and soon I think we're going to hit about 10,000 Bama tools in our cloud, which is massive and just to give you a scale off our network, our in French, our data center footprint. Our network introduction is about 30 network data centers that carry just network traffic across there are there across India and we're about eight application data centers across three regions. Data Center is like a five story building filled with servers. So we're talking really significant scale in India. And we had to do this because when we were launching, there are the government regulation and try it. They've gotten regulatory authority of India, mandates that any telecom company they have to store customer data inside India and none of the other cloud providers were big enough to host our clothes. Right. So we we made all this intellectual for ourselves, and we're still growing next. I love to show you how we grown with together with Moran says we started in 2014 with the fuel deployment pipelines, right? And then we went on to the NK deployment. Pipelines are cloud started growing. We started understanding the clouds and we picked up M C p, which has really been a game changer for us in automation, right on DNA. Now we are in the latest release, ofem CPM CPI $2019 to on open stack queens, which on we've just upgraded all of our clouds or the last few months. Couple of months, 2 to 3 months. So we've done about nine production clouds and there are about 50 internal, um, teams consuming cloud. We call as our tenants, right. We have open stack clouds and we have communities clusters running on top of open stack. There are several production grade will close that run on this cloud. The Geo phone, for example, runs on our cloud private cloud Geo Cloud, which is a backup service like Google Drive and collaboration service. It runs out of a cloud. Geo adds G o g S t, which is a tax filing system for small and medium enterprises, our retail post service. There are all these production services running on our private clouds. We're also empaneled with the government off India to provide cloud services to the government to any State Department that needs cloud services. So we were empaneled by Maiti right in their ego initiative. And our clouds are also Easter. 20,000 certified 20,000 Colin one certified for software processes on 27,001 and said 27,017 slash 18 certified for security processes. Our clouds are also P our data centers Alsop a 942 be certified. So significant effort and investment have gone toe These data centers next. So this is where I think we've really valued the partnership with Morantes. Morantes has has trained us on using the concepts of get offs and in fries cold, right, an automated deployments and the tool change that come with the M C P Morantes product. Right? So, um, one of the key things that has happened from a couple of years ago to today is that the deployment time to deploy a new 100 north production cloud has decreased for us from about 55 days to do it in 2015 to now, we're down to about five days to deploy a cloud after the bear metals a racked and stacked. And the network is also the physical network is also configured, right? So after that, our automated pipelines can deploy 100 0 clock in five days flight, which is a massive deal for someone for a company that there's adding bear metals to their infrastructure so fast, right? It helps us utilize our investment, our assets really well. By the time it takes to deploy a cloud control plane for us is about 19 hours. It takes us two hours to deploy a compu track and it takes us three hours to deploy a storage rack. Right? And we really leverage the re class model off M C. P. We've configured re class model to suit almost every type of cloud that we have, right, and we've kept it fairly generous. It can be, um, Taylor to deploy any type of cloud, any type of story, nor any type of compute north. Andi. It just helps us automate our deployments by putting every configuration everything that we have in to get into using infra introduction at school, right plus M. C. P also comes with pipelines that help us run automated tests, automated validation pipelines on our cloud. We also have tempest pipelines running every few hours every three hours. If I recall correctly which run integration test on our clouds to make sure the clouds are running properly right, that that is also automated. The re class model and the pipelines helpers automate day to operations and changes as well. There are very few seventh now, compared toa a few years ago. It very rare. It's actually the exception and that may be because off mainly some user letter as opposed to a cloud problem. We also have contributed auto healing, Prometheus and Manager, and we integrate parameters and manager with our even driven automation framework. Currently, we're using Stack Storm, but you could use anyone or any event driven automation framework out there so that it indicates really well. So it helps us step away from constantly monitoring our cloud control control planes and clothes. So this has been very fruitful for us and it has actually apps killed our engineers also to use these best in class practices like get off like in France cord. So just to give you a flavor on what stacks our internal teams are running on these clouds, Um, we have a multi data center open stack cloud, and on >>top of that, >>teams use automation tools like terra form to create the environments. They also create their own Cuba these clusters and you'll see you'll see in the next slide also that we have our own community that the service platform that we built on top of open stack to give developers development teams NGO um, easy to create an easy to destroy Cuban. It is environment and sometimes leverage the Murano application catalog to deploy using heats templates to deploy their own stacks. Geo is largely a micro services driven, Um um company. So all of our applications are micro services, multiple micro services talking to each other, and the leverage develops. Two sets, like danceable Prometheus, Stack stone from for Otto Healing and driven, not commission. Big Data's tax are already there Kafka, Patches, Park Cassandra and other other tools as well. We're also now using service meshes. Almost everything now uses service mesh, sometimes use link. Erred sometimes are experimenting. This is Theo. So So this is where we are and we have multiple clients with NGO, so our products and services are available on Android IOS, our own Geo phone, Windows Macs, Web, Mobile Web based off them. So any client you can use our services and there's no lock in. It's always often with geo, so our sources have to be really good to compete in the open Internet. And last but not least, I think I love toe talk to you about our container journey. So a couple of years ago, almost every team started experimenting with containers and communities and they were demand for as a platform team. They were demanding community that the service from us a manage service. Right? So we built for us, it was much more comfortable, much more easier toe build on top of open stack with cloud FBI s as opposed to doing this on bare metal. So we built a fully managed community that a service which was, ah, self service portal, where you could click a button and get a community cluster deployed in your own tenant on Do the >>things that we did are quite interesting. We also handle some geo specific use cases. So we have because it was a >>manage service. We deployed the city notes in our own management tenant, right? We didn't give access to the customer to the city. Notes. We deployed the master control plane notes in the tenant's tenant and our customers tenant, but we didn't give them access to the Masters. We didn't give them the ssh key the workers that the our customers had full access to. And because people in Genova learning and experimenting, we gave them full admin rights to communities customers as well. So that way that really helped on board communities with NGO. And now we have, like 15 different teams running multiple communities clusters on top, off our open stack clouds. We even handle the fact that there are non profiting. I people separate non profiting I peoples and separate production 49 p pools NGO. So you could create these clusters in whatever environment that non prod environment with more open access or a prod environment with more limited access. So we had to handle these geo specific cases as well in this communities as a service. So on the whole, I think open stack because of the isolation it provides. I think it made a lot of sense for us to do communities our service on top off open stack. We even did it on bare metal, but that not many people use the Cuban, indeed a service environmental, because it is just so much easier to work with. Cloud FBI STO provision much of machines and covering these clusters. That's it from me. I think I've said a mouthful, and now I love for you toe. I'd love to have your questions. If you want to reach out to me. My email is mine dot capulet r l dot com. I'm also you can also message me on Twitter at my uncouple. So thank you. And it was a pleasure talking to you, Andre. Let let me hear your questions.

>>from New York. It's the Q covering Escape. 19. >>Welcome back to the Cube coverage here in New York City for the first inaugural multi cloud conference called Escape, where in New York City was staying in New York, were not escaping from New York were in New York. It's all about multi Cloud, and we're here. Lisa Marie Nancy, developer advocate for Port Works, and Eric Conn, vice president of Products Works. Welcome back. Q. >>Thank you, John. Good to see >>you guys. So, um, whenever the first inaugural of anything, we want to get into it and find out why. Multi clouds certainly been kicked around. People have multiple clouds, but is there really multi clouding going on? So this seems to be the theme here about setting the foundation, architecture and data of the two kind of consistent themes. What shared guys take Eric, What's your take on this multi cloud trend? Yeah, >>I think it's something we've all been actively watching for a couple years, and suddenly it is becoming the thing right? So every we just had ah, customer event back in Europe last week, and every customer there is already running multi cloud. It's always something on their consideration. So there's definitely it's not just a discussion topic. It's now becoming a practical reality. So this event's been perfect because it's both the sense of what are people doing, What are they trying to achieve and also the business sense. So it's definitely something that is not necessarily mainstream, but it's becoming much more how they're thinking about building all their applications. Going forward, >>you know, you have almost two camps in the world. Want to get your thoughts on this guy's Because, like you have cloud native and people that are cloud native, they love it. They born the cloud that get it. Everything's cracking along. The developers air on Micro Service's They're agile train with their own micro service's. Then you got the hybrid I t. Trying to be hybrid developer, right? So you kind of have to markets coming together. So to me, I see multi cloud as kind of a combination of old legacy Data center types of I t with cloud native, not just ops and dead. But how about like trying to build developer teams inside enterprises? This seems to be a big trend, and multi club fits into that because now the reality is that I got azure. I got Amazon. Well, let's take a step back and think about the architecture. What's the foundation? So that to me, is more my opinion. But I want to get your thoughts and reactions that because if it's true, that means some new thinking has to come around around. What's the architecture? What are you trying to do? What's the workloads behavior outcome look like? What's the work flows? So there's a whole nother set of conversations that happened. >>I agree. I think the thing that the fight out there right now that we want to make mainstream is that it's a platform choice, and that's the best way to go forward. So it's still an active debate. But the idea could be I want to do multi club, but I'm gonna lock myself into the Cloud Service is if that's the intent or that's the design architecture pattern. You're really not gonna achieve the goals we all set out to do right, So in some ways we have to design ourselves or have the architecture that will let us achieve the business schools that were really going for and that really means from our perspective or from a port works perspective. There's a platform team. That platform team should run all the applications and do so in a multi cloud first design pattern. And so from that perspective, that's what we're doing from a data plan perspective. And that's what we do with Kubernetes etcetera. So from that idea going forward, what we're seeing is that customers do want to build a platform team, have that as the architecture pattern, and that's what we think is going to be the winning strategy. >>Thank you. Also, when you have the definition of cod you have to incorporate, just like with hybrid I t the legacy applications. And we saw that you throughout the years those crucial applications, as we call them People don't always want them to refer to his legacy. But those are crucial applications, and our customers were definitely thinking about how we're gonna run those and where is the right places it on Prem. We're seeing that a lot too. So I think when we talk about multi cloud, we also talk about what What is in your legacy? What is it? Yeah, I >>like I mean I use legacy. I think it's a great word because I think it really puts nail in the coffin of that old way because remember, if you think about some of the large enterprises, these legacy applications, they've been optimized for hardware and optimize their full stack. They've been build up from the ground up, so they're cool. They're running stuff, but it doesn't always translate to see a new platform designed point. So how do you mean Containers is great fit for their Cooper names. Obviously, you know is the answer. We you guys see that as well, but okay, I can keep that and still get this design point. So I guess what I want to ask you guys, as you guys are digging into some of the customer facing conversations, what are they talking about? The day talking about? The platform? Specifically? Certainly, on the security side, we're seeing everyone running away from buying tools to thinking about platforms. What's the conversation like on the cloud side >>way? Did a talk are multiplied for real talk at Barcelona? Q. Khan put your X three on Sudden. Andrew named it for reals of Izzy, but we really wanted to talk about multiplied in the real world. And when we said show of hands in Barcelona, who's running multi cloud? It was very, very few. And this was in, what, five months? Four months ago? Whereas maybe our customers are just really super advanced because of our 100 plus customers. At four words, we Eric is right. A lot of them are already running multi cloud or if not their plan, in the planning stage right now. So even in the last +56 months, this has become a reality. And we're big fans of communities. I don't know if you know Eric was the first product manager for Pernetti. Hey, he's too shy to say it on Dhe. So yeah, and we think, you know, and criminal justice to be the answer to making all They caught a reality right now. >>Well, I want to get back into G, K, E and Cooper. Very notable historic moment. So congratulations, But to your point about multi cloud, it's interesting because, you know, having multiple clouds means things, right? So, for instance, if I upgrade to office 3 65 and I kill my exchange server, I'm essentially running azure by their definition. If I'm building it, stack on AWS. I'm a native, this customer. Let's just say I want to do some tensorflow or play with big table or spanner on Google. Now >>we have three >>clouds now they're not. So they have work clothes, specific objectives. I am totally no problem. I see that like for the progressive customers, some legacy be to be people who like maybe they put their toe in the cloud. But anyone doing meaningful cloud probably has multiple clouds. But that's workload driven when you get into tying them together and is interesting. And I think that's where I think you guys have a great opportunity in this community because if open source convene the gateway to minimize the lock in and when I say lock and I mean like locking them propriety respect if his value their great use it. But if I want to move my data out of the Amazon, >>you brought up so many good points. So let me go through a few and Lisa jumping. I feel like locking. People don't wanna be locked >>in at the infrastructure level. So, like you said, if >>there's value at the higher levels of Stack, and it helps me do my business faster. That's an okay thing to exchange, but it is just locked in and it's not doing anything. They're that's not equal exchange, right, So there's definitely a move from infrastructure up the platform. So locking in >>infrastructure is what people are trying to move away from. >>From what we see from the perspective of legacy, there is a lot of things happening in industry that's pretty exciting of how legacy will also start to running containers. And I'm sure you've seen that. But containers being the basis you could run a BM as well. And so that will mean a lot for in terms of how V EMS can start >>to be matched by orchestrators like kubernetes. So that is another movement for legacy, and I wanted to acknowledge that point >>now, in terms of the patterns, there are definitely applications, like a hybrid pattern where connect the car has to upload all its data once it docks into its location and move it to the data center. So there are patterns where the workflow does move the ups are the application data between on Prem into a public cloud, for instance, and then coming back from that your trip with Lisa. There is also examples where regulations require companies to enterprise is to be able to move to another cloud in a reasonable time frame. So there's definitely a notion of Multi Cloud is both an architectural design pattern. But it's also a sourcing strategy, and that sourcing strategy is more regulation type o. R. In terms of not being locked in. And that's where I'm saying it's all those things. I'd >>love to get your thoughts on this because I like where you're going with this because it kind of takes it to a level of okay, standardization, kubernetes nights, containers, everyone knows what that is. But then you start talking about a P I gateways, for instance, right? So if I'm a car and I have five different gateways on my device, I ot devices or I have multiple vendors dealing with control playing data that could be problematic. I gotta do something like that. So I'm starting. Envision them? I just made that news case up, but my point is is that you need some standards. So on the a p I side was seeing some trends there. One saying, Okay, here's my stuff. I'll just pass parameters with FBI State and stateless are two dynamics. What do you make of that? What, What? What has to happen next to get to that next level of happiness and goodness? Because Bernays, who's got it, got it there, >>right? I feel that next level. I feel like in Lisa, Please jump. And I feel like from automation perspective, Kubernetes has done that from a P I gateway. And what has to happen next. There's still a lot of easy use that isn't solved right. There's probably tons of opportunities out there to build a much better user experience, both from the operations point of view and from what I'm trying to do is an intense because what people aren't gonna automate right now is the intent. They automate a lot of the infrastructure manual tasks, and that's goodness. But from how I docked my application, how the application did it gets moved. We're still at the point of making policy driven, easy to use, and I think there's a lot of opportunities for everyone to get better there. That's like low priority loving fruitcake manual stuff >>and communities was really good at the local food. That's a really use case that you brought up. Really. People were looking at the data now and when you're talking about persistent mean kun is his great for stateless, but for state full really crucial data. So that's where we really come in. And a number of other companies in the cloud native storage ecosystem come in and have really fought through this problem and that data management problem. That's where this platform that Aaron was talking to that >>state problem. Talk about your company. I want to go back to to, um, Google Days. Um, many war stories around kubernetes will have the same fate as map reduce. Yeah, the debates internally at Google. What do we do with it? You guys made the good call. Congratulations on doing that. What was it like to be early on? Because you already had large scale. You were already had. Borg already had all these things in place. Um, it wasn't like there was what was, >>Well, a few things l say one is It was intense, right? It was intense in the sense that amazing amount of intelligence amazing amount of intent, and right back then a lot of things were still undecided, right? We're still looking at how containers or package we're still looking at how infrastructure kit run and a lot of service is were still being rolled out. So what it really meant is howto build something that people want to build, something that people want to run with you and how to build an ecosystem community. A lot of that the community got was done very well, right? You have to give credit to things like the Sig. A lot of things like how people like advocates like Lisa had gone out and made it part of what they're doing. And that's important, right? Every ecosystem needs to have those advocates, and that's what's going well, a cz ah flip side. I think there's a lot of things where way always look back, in which we could have done a few things differently. But that's a different story for different. Today >>I will come back in the studio Palop of that. I gotta ask you now that you're outside. Google was a culture shock. Oh my God! People actually provisioning software provisioning data center culture shock when there's a little >>bit of culture shock. One thing is, and the funny thing is coming full circle in communities now, is that the idea of an application? Right? The idea of what is an application eyes, something that feels very comfortable to a lot of legacy traditional. I wanna use traditional applications, but the moment you're you've spent so much time incriminates and you say, What's the application? It became a very hard thing, and I used to have a lot of academic debates. Where is saying there is no application? It's It's a soup of resources and such. So that was a hard thing. But funny thing is covered, as is now coming out with definitions around application, and Microsoft announced a few things in that area to so there are things that are coming full circle, but that just shows how the movement has changed and how things are becoming in some ways meeting each other halfway. >>Talk about the company, what you guys are doing. Take a moment. Explain in context to multi cloud. We're here. Port works. What's the platform? It's a product. What's the value proposition? What's the state of the company. >>So the companies? Uh well, well, it's grown from early days when Lisa and I joined where we're probably a handful now. We're in four or five cities. Geography ease over 100 people over 150 customers and there. It's been a lot of enterprises that are saying, like, How do I take this pattern of doing containers and micro service is And how do I run it with my mission? Critical business crinkle workloads. And at that point, there is no mission critical business critical workload that isn't stable so suddenly they're trying to say, How do I run These applications and containers and data have different life cycles. So what they're really looking for is a data plane that works with the control planes and how controlled planes are changing the behavior. So a lot of our technology and a lot of our product innovation has been around both the data plane but a storage control plane that integrates with a computer controlled plane. So I know we like to talk about one control plane. There's actually multiple control planes, and you mentioned security, right? If I look at how applications are running way after now securely access for applications, and it's no longer have access to the data. Before I get to use it, you have to now start to do things like J W. T. Or much higher level bearer tokens to say, I know how to access this application for this life cycle for this use case and get that kind of resiliency. So it's really around having that storage. More complexity absolutely need abstraction >>layers, and you got compute. Look, leading work there. But you gotta have >>software to do it from a poor works perspective. Our products entirely software right down loans and runs using kubernetes. And so the point here is we make remarries able to run all the staple workloads out of the box using the same comment control plane, which is communities. So that's the experiences that we really want to make it so that Dev Ops teams can run anywhere close. And that's that's in some ways been part of the mix. Lisa, >>we've been covering Dev up, going back to 2010. Remember when I first was hanging around San Francisco 2008 joint was coming out the woodwork and all that early days and you look at the journey of how infrastructures code We talked about that in 2008 and now we'll get 11 years later. Look at the advancements you've been through this now The tipping point. It's just seems like this wave is big and people are on it. The developers air getting it. It's a modern renaissance of application developers, and the enterprises it's happening in the enterprise is not just like the nerds Tier one, the Alfa Geeks or >>the Cloud native. It's happening in the >>everyone's on board this time, and you and I have been in the trenches in the early stages of many open source projects. And I think with with kubernetes Arab reference of community earlier, I'm super proud to be running the world's largest CNC F for user group. And it's a great community, a diverse community, super smart people. One of my favorite things about working for works is we have some really smart engineers that have figured out what companies want, how to solve problems, and then we'll go creative. It'll open source projects. We created a project called autopilot, really largely because one of our customers, every who's in the G s space and who's running just incredible application. You can google it and see what the work they're doing. It's all there publicly, Onda We built, you know, we built an open source project for them to help them get the most out of kubernetes. We can say so. There's a lot of people in the community system doing that. How can we make communities better halfway make commitments, enterprise grade and not take years to do that? Like some of the other open source projects that we worked on, it took. So it's a super exciting time to be here, >>and open source is growing so fast now. I mean, just think about how these projects being structured. Maur and Maur projects are coming online and user price, but a lot more vendor driven projects to use be mostly and used, but now you have a lot of vendors who are users. So the line is blurring between Bender User in Open source is really fascinating. >>Well, you look at the look of the landscape on the C N. C f. You know the website. I mean, it's what 400 that are already on board. It's really important. >>They don't have enough speaking slasher with >>right. I know, and it's just it. It is users and vendors. Everybody's in this community together. It's one of things that makes it super exciting. And it it's how we know this is This was the right choice for us to base this on communities because that's what everybody, you guys >>are practically neighbors. So we're looking for seeing the studio. Palo Alto Eric, I want to ask you one final question on the product side. Road map. What you guys thinking As Kubernetes goes, the next level state, a lot of micro service is observe abilities becoming a key part of it, Obviously, automation, configuration management things are developing fast. State. What's the What's the road map for you guys? >>For us, it's been always about howto handle the mission critical and make that application run seamlessly. And then now we've done a lot of portability. So disaster recovery has been one of the biggest things for us is that customers are saying, How do I do a hybrid pattern back to your earlier question of running on Prem and in Public Cloud and do a d. R. Pale over into some of the things at least, is pointing out that we're announcing soon is non series autopilot in the idea, automatically managing applications scale from a volume capacity. And then we're actually going to start moving a lot more into some of the what you do with data after the life cycle in terms of backup and retention. So those are the things that everyone's been pushing us and the customers are all asking for. You >>know, I think data they were back in recovery is interesting. I think that's going to change radically. And I think we look at the trend of how yeah, data backup and recovery was built. It was built because of disruption of business, floods, our gains, data center failure. But I think the biggest disruptions ransomware that malware. So security is now a active disruptor. So it's not like it after the hey, if we ever have, ah, fire, we can always roll back. So you're infected and you're just rolling back infected code. That's a ransomware dream. That's what's going on. So I think data protection it needs to be >>redefined. What do you think? Absolutely. I think there's a notion of How do I get last week's data last month? And then oftentimes customers will say, If I have a piece of data volume and I suddenly have to delete it, I still need to have some record of that action for a long time, right? So those are the kinds of things that are happening and his crew bearnaise and everything. It gets changed. Suddenly. The important part is not what was just that one pot it becomes. How do I reconstruct everything? What action is not one thing. It's everywhere. That's right and protected all through the platform. If it was a platform decision, it's not some the cattlemen on the side. You can't be a single lap. It has to be entire solution. And it has to handle things like, Where do you come from? Where is it allowed to go? And you guys have that philosophy. We absolutely, and it's based on the enterprises that are adopting port works and saying, Hey, this is my romance. I'm basing it on Kubernetes. You're my date a partner. We make it happen. >>This speaks to your point of why the enterprise is in. The vendors jumped in this is what people care about Security. How do you solve this last mile problem? Storage. Networking. How do you plug those holes in Kubernetes? Because that is crucial to our >>personal private moment. Victory moment for me personally, was been a big fan of Cuban is absolutely, you know, for years. Then there were created, talked about one. The moments that got me that was really kind of a personal, heartfelt moment was enterprise buyer. And, you know, the whole mindset in the Enterprise has always been You gotta kill the old to bring in the new. And so there's always been that tension of a you know, the shiny new toy from Silicon Valley or whatever. You know, I'm not gonna just trash this and have a migration za paying that. But for I t, they don't want that to do that. They hate doing migrations, but with containers and kubernetes that could actually they don't to end of life to bring in the new project. They can do it on their own timetable or keep it around. So that took a lot of air out of the tension in on the I t. Side because they say great I can deal with the lifecycle management, my app on my own terms and go play with Cloud native and said to me, that's like that was to be like, Okay, there it is. That was validation. That means this Israel because now they can innovate without compromising. >>I think so. And I think some of that has been how the ecosystems embrace it, right. So now it's becoming all the vendors are saying my internal stack is also based on community. So even if you as an application owner or not realizing it, you're gonna take a B M next year and you're gonna run it and it's gonna be back by something like awesome. Lisa >>Marie Nappy Eric on Thank you for coming on Port Works Hot start of multiple cities Kubernetes big developer Project Open Source. Talking about multi cloud here at the inaugural Multi cloud conference in New York City. It's the Cube Courage of escape. 2019. I'm John Period. Thanks for watching

>>live from Atlanta, Georgia. It's the Q covering Answerable Fest 2019. Brought to you by >>Red Hat. >>Okay, welcome back. It runs two cubes. Live coverage of Ansel Fest here in Atlanta, Georgia. I'm John for a host of the Cube with stewed Minutemen. Analysts were looking angle. The Cube are next to guest Tom Anderson and most product owner. Red Hat is part of the sensible platform automation properly announced. And Richard Henshaw, product manager. Guys, welcome to the Cube Way had all the execs on yesterday and some customers all pretty jazzed up about this year, mainly around just the timing of how automation is really hitting the scene and some of the scale that's going on. You guys had big news with the answerable automation platform. New addition to the portfolio. What's the feedback? >>So far, I think the feedback has been super positive. We have customers have come to us. A lot of the last little one said, Hey, we're maturing. We're moving along the automation maturity curve, right, and we have multiple teams coming to us and saying, Hey, can you help us connect this other team? We've had a lot of success doing cloud provisioning or doing network automation were doing security automation. What have you and they're coming to us and saying, Help us give us kind of the story if you will, to be able to connect these other teams in our organization. And so that way I kind of feel the pole for this thing to move from a tool that automates this or that. This task for that task. Too much more of a platform center. >>It seems to be scaling out in terms of what automation is touching these days. And look at the numbers six million plus activations on get Hub versus other projects. So activities high in the community. But this seems to be much more broader. Scope now. Bring more things together. What's the rationale behind? What's the reasoning? What's the strategy? But the main thing is, >>automation is got to that point where it's becoming the skill set that we do. So it was always the focus. You know, I'm a database administrator. I'm assists out, man. I'm a middle where I'm a nap deaf on those people, then would do task inside their job. But now we're going to the point off, actually, anybody that can see apiece. Technology can automate piece technology in the clouds have shown This is the way to go forward with the things what we had. We bring that not just in places where it's being created from scratch, a new How do you bring that into what's existing? Because a lot of our customers have 20 or 30 years like a heritage in the I T estate. How do you do with all of that? You can't just rebuild everything into new as well. So you gotta be ableto automate across both of those areas and try and keep. You know, we say it's administrative efficiency versus organization effectiveness. Now how do I get to the point of the organization? Could be effective, supposed just doing things that make my job easier. And that's what we're gonna bring with applying automation capability that anybody can take advantage of. >>Richard. I actually felt the keynote demo this morning did a nice job of that line that they set it up with is this is this is tools that that all the various roles and teams just get it, and it's not the old traditional okay, I do my piece and set it up and then throw it over the wall. There was that, you know? Oh, I've got the notification and then some feedback loops and, you know, we huddled for something and it gets done rather fast, not magic. It's still when I get a certain piece done. Okay, I need to wait for it's actually be up and running, but you know, you're getting everybody into really a enterprise collaboration, almost with the tool driving those activities together >>on that. And that's why yesterday said that focus on collaboration is the great thing. All teams need to do that to be more successful because you get Maur inclusivity, Maurin puts. But organizations also need to coordinate what activities they're doing because they have rules, regulations, structures and standards they have to apply. Make sure that those people can do things in a way that's guided for them so that they're they're effective at what they're trying to do. >>Okay, I think I'm going to explain what's in the platform first because an engine and tower and there, what else is in there, what's new? What's what our customers is going to see. That's new. That's different >>it's the new components are automation Hope Collections, which is a technology inside answer ball itself. On also Automation Analytics and the casing is that engine and terrorist of the beating heart of the platform. But it's about building the body around the outside. So automation is about discover abilities like, What can we find out? What automation can I do that I'm allowed to do? Um, and let six is about the post activity. So I've automated all these things. I've done all this work well, How did it go? Who did what, who did? How much of what? How well did it work? How much did it failed? Succeeds and then, once you build on that, you don't start to expand out into other areas. So what? KP eyes, How much of what I do is automated versus no automated? You can start to instigate other aspects of business change, then Gamification amongst teams. Who's the Who's the boat? The closest motive here into the strategy input source toe How? >>Find out what's working right, essentially and sharing mechanism to for other groups in terms of knowing what's happening >>and how is my platform performing which areas are performing well, which airs might not be performing well. And then, as we move down the road, kind of how my performing against my peers are other organizations that are automating using the ants will automation platform doing? And am I keeping up on my doing better? That kind of stuff. >>So, Tom, there's a robust community as we was talking about. Their platform feels like it builds on yet to change the dynamic a little bit. When you talk about the automation hub and collections, you've already got a long list of the ecosystem vendors that are participating here. Bring us two through a little bit. What led Thio. You know all these announcements and where you expect, you know, how would this change the dynamics of >>the body? And maybe we'll split up that question. I'll talk a little bit about partners because it's both partners and customers in community here that's been driving us this way. I'll talk a little bit about partners and Rich talk about the customer piece here, which is partners have been traditionally distributing their content there. Ansel automation content through our engine capability. So our engine release cycle, or cadence, has been sort of the limiting factor to how fast they can get content out to their users and what what the collections does is part of the platforms allows us to separate those things. Rich talked about it yesterday in his keynote, having that stable platform. But you having yet having content be able to read fast. And our partners love that idea because they can content. They can develop content, create content, get into their users hands faster. So partners like at five and Microsoft you've seen on stage here are both huge contributors. And they've been part of the pole for us to get to the platform >>from a customer perspective. And the thing I love most about doing this job with the gas of customers is because I was a customer on Guy was danceable customer, and then I came over to this side on Dhe. I now go and see customers. I see what they've done, and I know what that's what I want to do. Or that's what I was trying to do. And she started to see those what people wanted to achieve, and I was said yesterday it is moving away from should I automate. How would we automate Maura? What should I automate? And so we'll start to see how customers are building their capabilities. And there's no there's many different ways people do. This is about different customers, >>you know. What's interesting is you guys have such a great success formula first. Well, congratulations. It's great to see how this is turning into such a wider market, because is not just the niche configuration management. More automation become with cloud to point a whole new wider category. So congratulations. The formula we see with success is good product, community customers adopting and then ecosystem that seems to be the successful former in these kinds of growth growth waves you guys experiencing? What is the partnering with you mentioned? S five Microsoft? Because that, to me, is gonna be a tipping point in a tel sign for you guys because you got the community. You got the customers that check check ecosystem. What's the partner angle? How do they involve? Take us through that. What's going on? They're >>so you're absolutely so you know, kind of platform velocity will be driven by partner adoption and how many things customers can automate on that platform or through that platform and for us I mean, the example was in the demo this morning where they went to the automation hub and they pulled down the F five collection, plugged it into a workflow, and they were automating. What are partners? Experience through their customers is Look, if I'm a customer, I have a multi cloud environment or hybrid cloud environment. I've got automation from AWS. I've got azure automation via more automation. Five. Got Sisko. I've got Palo Alto. I've got all these different automation tools to try and string them together, and the customers are coming and telling those vendors Look, we don't want to use your automation to end this automation tooling that one we want to use Ansel is the common substrate if you will automation substrate across this platform. So that's motivating the partners to come to us and say, Hey, I had I was out five Aspire last week, and they're all in a natural. I mean, it's really impressive to see just how much there in unanswerable and how much they're being driven by their customers when they do Ansell workshops without five, they say the attendance is amazing so they're being pulled by their customers and therefore the partners are coming to us. And that's driving our platform kind of usability across the across the scale. >>Another angle we'll see when we talk to the engineers of the partners that are actually doing the work to work with danceable is that they're seeing is ah, change also in how they it's no longer like an individual customer side individual day center because everything is so much more open and so much more visible. You know there's value in there, making it appealing and easy for their customers to gain advantage of what they're doing. And also the fact that the scales across those customers as well because they have their internal team's doing it, saying the same things and so bringing them to an automation capable, like Ansel have to push. That means that they also gained some of the customers appreciation for them, making it easier to do their tasking collaboration with us and you know, the best collaborations. We've got some more partners, all initiated by customers, saying Hey, I want you to go and get danceable content, >>the customer driving a lot of behavior, the guest system. Correct. On the just another point, we've been hearing a lot of security side separate sector, but cyber security. A lot of customers are building teams internally, Dev teams building their own stacks and then telling the suppliers a support my AP eyes. So now you start to see more of a P I integration point. Is that something that is gonna be something that you guys gonna be doubling down on? What's that? What's the approach there? How does that partner connected scale with the customers? So we've >>been eso Ansel security automation, which is the automation connecting I. P. S. C. P. S that kind of stuff. It is almost a replay of what we did the network automation space. So we saw a need in the network automation space. We feel that we became a catalyst in the community with our partners and our customers and our and our contributors. And after about three years now, Ansel Network automation is a huge piece of our business and adoption curve. We're doing the exactly see the exact same thing in the security automation space compliance. The side over here, we're talking about kind of automating the connections between your firewalls, your threat detection systems and all that kind of stuff. So we're working with a set of partners, whether it's Cisco, whether it's Palo Alto, whether it's whether it's resilient by the EMS, resilient and being able to connect and automate the connections between the threat and the response and and all of that kind of >>the same trajectory as the network automation >>Zach. Same trajectory, just runnin the same play and it's working out right now. We're on that kind of early part of that curve, that adoption curve, and we have partners jumping in with us. >>You're talking to customers. We've heard certain stories. You know how I got, you know, 1000 hours of work down to a dozen hours of work there. Is there anything built into the tool today that allows them to kind of generate those those hero stats O. R. Any anything along those lines? >>Talk about analytic committee from yes, >>well, again without any analytic side. I mean, those things starts become possible that one of the things we've been doing is turning on Maur more metrics. And it's actually about mining the data for the customer because Tower gives this great focal point for all the automation that's going on. It's somewhere that everything comes through. So when we export that and then we can we can do that work for all the customers rather than have to duel themselves. Then you start to build those pictures and we start with a few different areas. But as we advance with those and start, see how people use them and start having that conversation customers about what data they want to use and how they want to use it, I think that's gonna be very possible. You know, it's so >>important. E think was laid out here nicely. That automation goes from a tactical solution to more strategic, but more and more how customers can leverage that data and be data driven. That's that's gonna drive them for it. And any good customer examples you have of the outcomes. No, you're talking to a lot of >>PS one from this morning. Yeah, >>so I mean, I'll be Esther up this morning, and I think that the numbers they used in the demo that she's like, you know, last year they did 100,000 from launch to the end of the year. 100,000 changes through their platform on this year so far that in a 1,000,000. So now you know, from my recollection, that's about the same time frame on either side of the year. So that's a pretty impressive acceleration. Side of things. We've had other ones where people have said, You know how many times you were telling some customers yesterday? What used to take eight hours to a D R test with 20 or 30 people in for the weekend now takes 12 minutes for two People on the base is just pushing a few buttons just as they go through and confirm everything worked that that type of you can't get away from that type of change. >>J. P. Morgan example yesterday was pretty compelling. I mean, time savings and people are, I mean, this legit times. I mean, we're talking serious order of magnitude, time savings. So that's awesome. Then I want to ask you guys, Next is we're seeing another pattern in the market where amongst your customer base, where it's the same problem being automated, allover the place so playbooks become kind of key as that starts to happen is that where the insights kind of comes in? Can you help us kind of tie that together? Because if I'm a large enterprise with its I'm decentralized or centralized, are organized problem getting more gear? I'm getting more clouds, game or operations. There's more surface area of stuff and certainly five g I ot is coming around the corner. Mention security. All this is expanding to be much more touchpoints. Automation seems to be the killer app for this automation, those mundane task, but also identifying new things, right? Can you guys comment on that? >>Yeah, so maybe I'll start rich. You could jump in, which is a little bit around, uh, particularly those large accounts where you have these different disparate teams taking a approach to automate something, using Ansel and then be able to repeat or reuse that somewhere else. The organization. So that idea of being for them to be able to curate they're automation content that they've created. Maybe they pulled something down from galaxy. Maybe they've got something from our automation husband. They've made it their own, and now they want to curate that and spread it across the organization to either obviously become more efficient, but also in four standards. That's where automation hub is going to come into play here. Not only will it be a repo for certify content from us and our partners, but it will also be an opportunity for them to curate their own content and share it across the organization. >>Yeah, I think when you tie those two things together and you've got that call discover abilities, I had away go and find what I want. And then the next day, the next day, after you've run the automation, you then got the nerve to say, Well, who's who's using the right corporate approved rolls? Who's using the same set of rolls from the team that builds the standards to make sure you're gonna compliant build again, showing the demo That's just admin has his way of doing it, puts the security baseline application on top and you go, Oh, okay, who's running that security baseline continuously every time. So you can both imposed the the security standards in the way the build works. But you can also validate that everybody is actually doing the security standards. >>You what I find fascinating about what you guys are doing, and I think this is came out clearly yesterday and you guys are talking about it. And some of the community conversations is a social construct here. Going on is that there's a cultural shift where the benefits that you guys are throwing off with the automation is creating a network effect within the companies. So it's not just having a slack channel on texting. The servers are up or down. It's much more of a tighter bond between the stakeholders inside the company's. Because you have people from different geography is you have champions driving change. And there's some solidarity happening between the groups of people, whether they're silo door decentralized. So there's a whole new social network, almost a cultural shift that's happening with the standardization of the substrate. Can you guys comment on this dynamic? Did you see this coming? You planning forward? Are you doubling down on it? >>I think so. And we talk about community right on how important that is. But how did you create that community internally and so ask balls like the catalyst so most teams don't actually need to understand in their current day jobs. Get on all the Dev ops, focus tools or the next generation. Then you bring answer because they want to automate, and suddenly they go. Okay, Now I need to understand source control, and it's honest and version. I need to understand how to get pulls a full request on this and so on and so forth on it changes that provides this off. The catalyst for them to focus on what changed they have to make about how they work, because what they wanted to do was something that requires them to do you no good disciplines and good behaviors that previously there was no motivation or need to do. I think >>Bart for Microsoft hit on that yesterday. You know, if you saw Bart Session but their network engineers having to get familiar with concepts of using automation almost like software development, life cycles right and starting to manage those things in repose. And think of it that way, which is intimidating at first for people who are not used to. But once they're over that kind of humping understand that the answer language itself is simple, and our operations person admin can use it. No problem, >>he said himself. Didn't my network engineers have become network developers. >>It's funny watching and talking to a bunch of customers. They all have their automation journey that they're going through. And I hear the Gamification I'm like, Okay, what if I have certain levels I have to reach in it unlocked capabilities, you know, in the community along the way. Maybe that could build a built in the future. >>Maybe it's swag based, you know, you >>get level C shows that nice work environment when you're not talking about the server's down on some slack channel when you're actually focusing on work. Yeah, so that mean that's the shift. That's what I'm saying, going >>firefighting to being able to >>do for throwing bombs. Yeah, wars. And the guy was going through this >>myself. Now you start a lot of the different team to the deaf teams and the ops teams. And I say it would be nice if these teams don't have to talk to complain about something that hadn't worked. It was Mexican figured it was just like I just like to talk to you because you're my friend. My colleague and I'd like to have a chat because everything's working because it's all automated, so it's consistent. It's repeatable. That's a nice, nice way. It can change the way that people get to interact because it's no longer only phoned me up when something's wrong. I think that absent an interesting dynamic >>on our survey, our customer base in our community before things one of the four things that came up was happier employees. Because if they're getting stuff done and more efficient, they have more time to actually self actualizing their job. That becomes an interesting It's not just a checkbox in some HR manual actually really impact. >>And I kind of think the customers we've heard talk rvs, gentlemen, this morning gave me a lot of the fear initially is, well, I automate myself out of a job, and what we've heard from everybody is that's not absolutely That's not actually true at all. It just allows them to do higher value things that, um or pro >>after that big data, that automation thing. That's ridiculous. >>I didn't use it yesterday. My little Joe Comet with that is when I tried to explain to my father what I do. Andi just said Well, in the 19 seventies, they said that computers you mean we'll do a two day week on? That hasn't come >>true. Trade your beeper and for a phone full of pots. But Richard, Thanks for coming on. Thanks for unpacking the ants. Full automation platforms with features. Congratulations. Great to see the progress. Thank you, Jonah. Everybody will be following you guys to Cuba. Coverage here in Atlanta, First Amendment Stevens for day two of cube coverage after this short break.

>>live from Atlanta, Georgia. It's the Q covering Answerable Fest 2019. Brought to you by Red Hat. >>Hey, welcome back, everyone. It's the cubes. Live coverage here in Atlanta, Georgia, for answerable fast. Part of redheads. Big news. Ansel Automation Platform was announced. Among other things, they're great products. I'm John for ear, with my coast to minimum, but two great guests. You unpack all the automation platform features and benefits. Walter Bentley, senior manager. Automation Practicing red hat and vj Job Olu, director of Red Hat Consulting Guys Thanks for coming on. Thanks. So the activity is high. The buzz this year seems to be at an inflection point as this category really aperture grows big time seeing automation, touching a lot of things. Standardization. We heard glue layer standard substrate. This is what answer is becoming so lots of service opportunity, lot of happy customers, a lot of customers taking it to the next level. And a lot of customers trying to consolidate figure out hadn't make answerable kind of a standard of other couples coming in. You guys on the front lines doing this. What's the buzz? What's the main store? What's the top story going on around the service is how to deploy this. What are you guys seeing? >>So I think what we're seeing now is customers. Reactor building automation. For a long time, I have been looking at it at a very tactical level, which is very department very focused on silo. Whether country realizes with this modern develops and the change in how they actually go to the market, they need to bring the different teams together. So they're actually looking at watching my enterprise automation strategy be how to actually take what I've learned in one organization. And I still roll it across the enterprise so that now struggling and figuring out how to be scared, what we have, how do we change the culture of the organization to collaborate a lot more and actually drive automation across enterprise? >>Walter One of the things we've been we've talked about all the time in the Cube, and it's become kind of cliche. Digital transformation. Okay, I heard that before, and three things people process, technology, process and capability you guys have done You mentioned the siloed having capabilities that's been there. Check was done very, very well as a product technology Red hat in the portfolio. Great synergies. We talked about rail integration, all the benefits there. But the interesting thing this year that I've noticed is the people side of the equation is interesting. The people are engaged, is changing their role because automation inherently changes there, function in the organization because it takes away probably the mundane tasks. This is a big part of the equation. You guys air hitting that mark. How do you How are you guys seeing that? How you accelerating that has that changing your job, >>right? So customers are now economy realizing that going after automation in a very tactical manner is not exactly getting them what they want as a far as a return on investment in the automation. And what they're realizing is that they need to do more. And they're coming to us and more of an enterprise architectural level and say we want to talk mortgage grander strategy. And what they're coming to realize is that having just one small team of people that were calling the Dev Ops team is not gonna be ableto drive that adoption across the organization. So what we're trying to do is work with customers to show them how they collaboration in the culture of peace is huge. It's a huge part of adopting automation. Answerable is no longer considered a emerging tech anymore. And and I when I say that, I mean a lot of organizations are using answerable in many different ways. They're past that point, and now they're moving on to the next part, which is what is our holistic strategy and how we're gonna approach automation. And And we wanted leverage danceable, unanswerable tower to do that. >>Does that change how you guys do your roll out your practices in some of your programs? >>Well, we did have to make some adjustments in the sense of recognizing that the cultural piece is a pivotal part of it, and we can go in and we can write playbooks and rolls, and we can do all those things really great. But now we need to go in and help them structure themselves in a way where they can foster that collaboration and keep a moment. >>And I'll actually add on to that so reactive, large, open innovation labs three years ago, and what we have to learn doing that is using labs and allows practices to actually help customers embrace new culture and change. How they actually operate has actually helped us take those practices and bring it into our programs and kind of drive that to our customers. So we actually run our automation adoption program and the journey for customers through those practices that we actually learned in open innovation loves like open practice, library, even storming priority sliders and all of those modern techniques. So the goal is to help our customers understand those practices and actually embrace them and bring them into the organization to drive the change that that's looking for within the organization. >>A. J. Is there anything particular for those adoption practices when you're talking about Cloud? Because the communication amongst teams silos, you know, making things simpler is something that we absolutely do need for cloud. So I'm just curious how you connect kind of the cloud journey with the automation journey. >>So all of the journey program that actually created, whether it's a contender adoption program or the automation adoption program, we actually followed the same practices. So whether you're actually focused on a specific automation to, like, answerable or actually embarking on hybrid multicolored journey. We actually use the same practices so the customers don't have toe learn new things every time you have to go from one product, one of the so that actually brings a consistent experience to customers in driving change within the organization. So let's picture whether it is focusing automation focused on cloud migrating to the cloud. The practices remained the same, and the focus is about not trying to boil the ocean on day one. Try to break it into manageable chunks that give it a gun back to the business quickly learned from the mistakes that you make in each of the way and actually build upon it and actually be successful. >>So, Walter, I always love when we get to talk to the people that are working straight with customers because you come here to the conference, it's like, Oh, it's really easy Get started. It doesn't matter what role or what team you're in. Everybody could be part of it. But when you get to the actual customers, they're stumbling blocks. You know what are some of those things? What are some of the key things that stop people from taking advantage of all the wonderful things that all the users here are doing >>well. One of the things that I've identified and we've identified as a team is a lot of organizations always want to blow the ocean. And when and when it comes down to automation, they feel that if they are not doing this grand transformation and doing this this huge project, then they're not doing automation. And the reality is is that we're Trent with showing them that you can break things up into smaller chunks, as Visi alluded to. And even if you fail, you fail fast and you can start over again because you're dealing with things in a smaller chunk. And we've also noticed that by doing that, we're able to show them to return on investment faster so they can show their leadership, and their leadership can stand behind that and want to doom. Or so that's one of the areas. And then I kind of alluded to the other area, which is you have to have everybody involved. You want just subject matter experts riding content to do the automation. You don't want that just being one silo team. You want to have everybody involved and collaborate as much as possible. >>Maybe can you give us an example? Is about the r A y How fast to people get the results and, you know, prove toe scale this out. >>So with the automation adoption journey, what we're able to do is is that we come in and sit down with our customers and walk them through how to properly document their use cases. What the dependencies, What integration points, possibly even determining what is that? All right, ranking for that use case. And then we move them very quickly in the next increment. And in the next increment, we actually step them through, taking those use cases, breaking them down into minimum viable products and then actually putting those in place. So within a 90 day or maybe a little bit more than a little bit more than the 90 day window, were able to show the customer in many different parts of the organization how they're able to take advantage of automation and how the return on investment with hopes of obviously reducing either man hours or being able to handle something that is no a mundane task that you had to do manually over and over again. >>What are some of the things that people get confused about when they look at the breath of what's going on with the automation platform? When I see tool to platform, transitions are natural. We've seen that many times in the industry that you guys have had product success, got great community, that customers, they're active. And now you've got an ecosystem developing so kind of things air popping on all cylinders here. >>So the biggest challenge that we're actually being seeing customers is they actually now come to realize that it's very difficult to change the culture of the organization right there, actually embarking on this journey and the biggest confusion that is, how do we actually go make those changes? How do we bring some of the open practice some of the open source collaboration that Riddle had into the organization so they actually can operate in a more open source, collaborative way, and what we have actually learned is we actually have what we call its communities of practice within Red Hack. It is actually community off consultants, engineers and business owners. The actual collaborate and work together on offering the solutions to the market. So we're taking those experiences back to our customers and enabling them to create those communities of practice and automation community that everybody can be a part off. They can share experiences and actually learn from each other much easier than kind of being a fly on the wall or kind of throwing something or defense to see what sticks and what does not. >>What's interesting about the boiling the ocean comment you mentioned Walter and B J is your point. There is, is that the boil? The ocean is very aspirational. We need change rights. That's more of the thing outcome that they're looking for. But to get there is really about taking those first steps, and the folks on the front lines have you their applications. They're trying to solve or manage. Getting those winds is key. So one of things that I'm interested in is the analytics piece showing the victory so in the winds early is super important because that kind of shows the road map of what the outcome may look like versus the throw the kitchen, sink at it and, you know, boil the ocean of which we know to the failed strategy. Take us through those analytics. What are some of the things that people tend to knock down first? What are some of the analytical points that people look at for KP eyes? Can you share some insight into that? >>Sure, sure. So we always encourage our customers to go after the platform first. And I know that may sound the obvious, but the platform is something that is pretty straightforward. Every organization has it. Every organization struggles with provisioning, whether of a private cloud, public cloud, virtualization, you name it. So we have the customer kind of go after the platform first and look at some of their day to operations. And we're finding that that's where the heaviest return on investment really sits. And then once you get past that, we can start looking like in the end, work flows. You know, can they tie service now to tower, to be able to make a complete work flow of someone that's maybe requesting a BM, and they can actually go through that whole workflow by by leveraging tower and integration point like service. Now those air where we're finding that the operators of these systems going getting the fastest benefit. And it also, of course, benefits the business at the end of the day because they get what they need a lot fast. >>It's like a best practice and for you guys, you've seen that? Yes, sir. Docked with that out of E. J. What's your comment on all this? >>So going back to the question on metrics Automation is great, but it does not provide anybody to the business under the actually show. What was the impact, whether it's from a people standpoint, cost standpoint or anything else. So what we try to drive is enable customers. You can't build the baseline off where they are today, and as they're going through the incremental journey towards automation, measure the success of that automation against the baseline. And that actually adds the other way back to the customer. As a business you didn't get to see. I was creating a storage land. I was doing it probably 15 times a month. Take it or really even automated. It spend like a day created a playbook. I'll save myself probably half, of course, and that could be doing something that's better. So building those metrics and with the automation analytics that actually came in the platform trying those bass lines. So the number of executions, actually the huge value they'll actually be ableto realize the benefits of automation and measure the success off within enterprise. >>So I'm a customer prospect, like I want to get a win. I don't want to get fired. I won't get promoted. Right, I say, Okay, I gotta get a baseline and knock down some playbooks. Knock that down first. That what you're gonna getting it. That's a good starting. >>Starting. Understand your baseline today. Plan your backlog as to what you want to knock down. And once you know them down, build a dashboard as to what the benefits were, what the impact was actually built upon it. You actually will see an incremental growth in your success with automation. >>And then you go to the workflow and too, and that's your selling point for the next level. Absolutely good playbook. Is that the automation programs that in a nutshell or is that more of a best practice >>those components of the ah, the automation adoption journey that we allow the customer to kind of decide how they want their journey to be crafted. Of course, we have a very specific way of going about and walking them through it. But we allowed in the kind of crap that journey and that is those the two components that make up the automation. >>We're gonna put you guys on the spot with the tough question We heard from G. P. Morgan yesterday on the Kino, which I thought was very compelling. You know, days, hours, two minutes. All this is great stuff. It's real impact. Other customers validate that. So, congratulations. Can you guys share any anecdotal stories? You know, the name customers? Just about situations Where customs gone from this to this old way, new way and throw some numbers around Shearson Samantha >>is not a public reference, but I like to give you a customer. Exactly. Retail company. When we first actually went and ran a discovery session, it took them 72 days to approach in an instance. And the whole point was not because it took that long. It because every task haven't s l. A We're actually wait for the Acela manually. Go do that. We actually went in >>with our 72 hours, two days, two days, >>actually, going with the automation? We Actually, it was everybody was working on the S L. A. We actually brought it down to less than a day. So you just gave the developers looking to code 71 days back for him to start writing code. So that's the impact that we see automation bringing back to the customers, right? And you'll probably find the use causes across everywhere. Whether J. P. Morgan Chase you actually had the British Army and everyone here on states talking about it. It is powerful, but it is powerful relief you can measure and learn from it >>as the baseline point. Get some other examples because that's that's, uh, that's 70 days is that mostly delay its bureaucracy. It's It's so much time. >>It's manual past and many of the manual tasks that actually waiting for a person to do the task >>waterfall past things sound, although any examples you can >>yes, so the one example that always stands out to me and again, it's a pretty interviewing straight forward. Is Citrix patching? So we work with the organization. They were energy company, and they wanted to automate patching their searches environment, patching this citrus environment took six weekends and it took at least five or six engineers. And we're talking about in bringing an application owners, the folks who are handling the bare metal, all all that whole window. And by automating most of the patching process, we were able to bring it down to one weekend in one engineer who could do it from home and basically monitor the process instead of having to be interactive and active with it. And to me, that that was a huge win. Even though it's, you know, it's such dispatching. >>That's the marketing plan. Get your weekends back. Absolutely awesome. Shrimp on the barbecue, You know, Absolutely great job, guys. Thanks for the insight. Thanks. Come on. The key. Really appreciate it. Congratulations. Thank you. Thanks for sharing this queue here. Live coverage. Danceable fest. Where the big news is the ass. Full automation platform. Breaking it down here on the Q. I'm John. First to Minutemen. We're back with more coverage after this short break

>> from our studios in the heart of Silicon Valley, Palo Alto, California It is a cute conversation. >> Well, the Special Cube conversation. We are here in Palo Alto, California, Cube studios here. Tony, Gino, Domenico, Who's the senior security strategist and research at for Net and four to guard labs live from Las Vegas. Where Black Hat and then Def Con security activities happening, Tony, also known as Tony G. Tony G. Welcome to this cube conversation. >> Hey, Thanks, John. Thanks for having me. >> So a lot of action happening in Vegas. We just live there all the time with events. You're there on the ground. You guys have seen all the action there. You guys are just published. Your quarterly threat report got a copy of it right here with the threat index on it. Talk about the quarterly global threats report. Because the backdrop that we're living in today, also a year at the conference and the cutting edge is security is impacting businesses that at such a level, we must have shell shock from all the breaches and threats they're going on. Every day you hear another story, another story, another hack, more breaches. It said all time high. >> Yeah, you know, I think a lot of people start to get numb to the whole thing. You know, it's almost like they're kind of throwing your hands up and say, Oh, well, I just kind of give up. I don't know what else to do, but I mean, obviously, there are a lot of different things that you can do to be able to make sure that you secure your cybersecurity program so at least you minimize the risk of these particular routes is happening. But with that said with the Threat Landscape report, what we typically dio is we start out with his overall threat index, and we started this last year. If we fast forward to where we are in this actual cue to report, it's been one year now, and the bad news is that the threats are continuing to increase their getting more sophisticated. The evasion techniques are getting more advanced, and we've seen an uptick of about 4% and threat volume over the year before. Now the silver lining is I think we expected the threat volume to be much higher. So I think you know, though it is continuing to increase. I think the good news is it's probably not increasing as fast as we thought it was going to. >> Well, you know, it's always You have to know what you have to look for. Blood. People talk about what you can't see, and there's a lot of a blind spot that's become a data problem. I just want to let people know that. Confined the report, go to Ford Nets, ah website. There's a block there for the details, all the threat index. But the notable point is is only up 4% from the position year of a year that the attempts are more sophisticated. Guys gotta ask you, Is there stuff that we're not seeing in there? Is there blind spots? What's the net net of the current situation? Because observe ability is a hot topic and cloud computing, which essentially monitoring two point. Oh, but you gotta be able to see everything. Are we seeing everything? What's what's out there? >> Well, I mean, I think us as Ford, a guard on Darcy, have cyber threat in challenges. I think we're seeing a good amount, but when you talk about visibility, if you go back down into the organizations. I think that's where there's There's definitely a gap there because a lot of the conversations that I have with organizations is they don't necessarily have all the visibility they need from cloud all the way down to the end point. So there are some times that you're not gonna be able to catch certain things now. With that said, if we go back to the report at the end of the day, the adversaries have some challenges to be able to break into an organization. And, of course, the obvious one is they have to be able to circumvent our security controls. And I think as a security community, we've gotten a lot better of being able to identify when the threat is coming into an organization. Now, on the flip side, Oh, if you refer back to the minor Attack knowledge base, you'll see a specific tactic category called defense evasions. There's about 60 plus techniques, evasion techniques the adversary has at their disposal, at least that we know may there may be others, but so they do have a lot of opportunity, a lot of different techniques to be able to leverage with that, said There's one technique. It's, ah, disabling security tools that we started seeing a bit of an increase in this last cue to threat landscape report. So a lot of different types of threats and mile where have the capability to be ableto one look at the different processes that may be running on a work station, identifying which one of those processes happen to be security tools and then disabling them whether they're no, maybe they might just be able to turn the no, the actual service off. Or maybe there's something in the registry that they can tweak. That'll disable the actual security control. Um, maybe they'll actually suppress the alerts whatever. They conduce you to make sure that that security control doesn't prevent them from doing that malicious activity. Now, with that said, on the flip side, you know, from an organization for perspective, you want to make sure that you're able to identify when someone's turning on and turning off those security control to any type of alert that might be coming out of that control also. And this is a big one because a lot of organizations and this certainly do this minimize who has the ability to turn those particular security controls on and off. In the worst cases, you don't wanna have all of your employees uh, the you don't want to give them the ability to be able to turn those controls on and off. You're never gonna be ableto baseline. You're never gonna be able to identify a, you know, anomalous activity in the environment, and you're basically gonna lose your visibility. >> I mean, this increase in male wearing exploit activity you guys were pointing out clearly challenge the other thing that the report kind of She's out. I want to get your opinion on this. Is that the The upping? The ante on the evasion tactics has been very big trend. The adversaries are out there. They're upping the ante. You guys, we're upping the guarantees. This game you continue this flight will continues. Talk about this. This feature of upping the ante on evasion tactics. >> Yes. So that's what I was that I was kind of ah, referring to before with all the different types of evasion techniques. But what I will say is most of the all the threats these days all have some type of evasion capabilities. A great example of this is every quarter. If you didn't know. We look at different types of actors and different types of threats, and we find one that's interesting for us to dig into and where create was called an actual playbook, where we want to be able to dissect that particular threat or those threat actor methodologies and be able to determine what other tactics and corresponding techniques, which sometimes of course, includes evasion techniques. Now, the one that we focused on for this quarter was called His Ego's Was Ego, says a specific threat that is an information stealer. So it's gathering information, really based on the mission goals off, whatever that particular campaign is, and it's been around for a while. I'm going all the way back to 2011. Now you might be asking yourself, Why did we actually choose this? Well, there's a couple different reasons. One happens to be the fact that we've seen an uptick in this activity. Usually when we see that it's something we want to dive into a little bit more. Number two. Though this is a tactic of the of the adversary, what they'll do is they'll have their threat there for a little while, and then local doorman. They'll stop using that particular malware. That's no specific sort of threat. They'll let the dust settle that things die down. Organizations will let their guard down a little bit on that specific threat. Security organizations Ah, vendors might actually do the same. Let that digital dust kind of settle, and then they'll come back. Bigger, faster, stronger. And that's exactly what Z ghosted is. Ah, we looked at a specific campaign in this new mall where the new and improved Mauer, where is they're adding in other capabilities for not just being able to siphon information from your machine, but they're also now can capture video from your webcam. Also, the evasion techniques since Iran that particular subject, what they're also able to do is they're looking at their application logs. Your system logs your security logs, the leading them making a lot more difficult from a forensic perspective. Bill, go back and figure out what happened, what that actual malware was doing on the machine. Another interesting one is Ah, there. We're looking at a specific J peg file, so they're looking for that hash. And if the hash was there the axle? Um, our wouldn't run. We didn't know what that was. So we researched a little bit more on What we found out was that J Peg file happened to be a desktop sort of picture for one of the sandboxes. So it knew if that particular J pick was present, it wasn't going to run because it knew it was being analyzed in a sandbox. So that was a second interesting thing. The 3rd 1 that really leaned us towards digging into this is a lot of the actual security community attribute this particular threat back to cyber criminals that are located in China. The specific campaign we were focused on was on a government agency, also in China, So that was kind of interesting. So you're continuing to see these. These mile wears of maybe sort of go dormant for a little bit, but they always seem to come back bigger, faster, stronger. >> And that's by design. This is that long, whole long view that these adversaries we're taking in there as he organized this economy's behind what they're doing. They're targeting this, not just hit and run. It's get in, have a campaign. This long game is very much active. Howto enterprises. Get on, get on top of this. I mean, is it Ah, is it Ah, people process Issue is it's, um, tech from four to guard labs or what? What's what's for the Nets view on this? Because, I mean, I can see that happening all the time. It has >> happened. Yeah, it's It's really it's a combination of everything on this combination. You kind of hit like some of it, its people, its processes and technology. Of course, we have a people shortage of skilled resource is, but that's a key part of it. You always need to have those skills. Resource is also making sure you have the right process. Is how you actually monitoring things. I know. Ah, you know, a lot of folks may not actually be monitoring all the things that they need to be monitoring from, Ah, what is really happening out there on the internet today? So making sure you have clear visibility into your environment and you can understand and maybe getting point in time what your situational awareness is. You you, for my technology perspective, you start to see and this is kind of a trend. We're starting the leverage artificial intelligence, automation. The threats are coming, and it's such a high volume. Once they hit the the environment, instead of taking hours for your incident response to be about, at least you know not necessarily mitigate, but isolate or contain the breach. It takes a while. So if you start to leverage some artificial intelligence and automatic response with the security controls are working together. That's a big that's a big part of it. >> Awesome. Thanks for coming. This is a huge problem. Think no one can let their guard down these days? Certainly with service, they're expanding. We're gonna get to that talk track in the second. I want to get quickly. Get your thoughts on ransom, where this continues to be, a drum that keeps on beating. From a tax standpoint, it's almost as if when when the attackers need money, they just get the same ransomware target again. You know, they get, they pay in. Bitcoin. This is This has been kind of a really lucrative but persistent problem with Ransomware. This what? Where what's going on with Ransomware? What's this state of the report and what's the state of the industry right now in solving that? >> Yeah. You know, we looked into this a little bit in last quarter and actually a few quarters, and this is a continuous sort of trend ransom, where typically is where you know, it's on the cyber crime ecosystem, and a lot of times the actual threat itself is being delivered through some type of ah, phishing email where you need a user to be able to click a langur clicking attachment is usually kind of a pray and spray thing. But what we're seeing is more of ah, no sort of ah, you know, more of a targeted approach. What they'll do is to look for do some reconnaissance on organizations that may not have the security posture that they really need. Tohave, it's not as mature, and they know that they might be able to get that particular ransomware payload in there undetected. So they do a little reconnaissance there, And some of the trend here that we're actually seeing is there looking at externally RTP sessions. There's a lot of RTP sessions, the remote desktop protocol sessions that organizations have externally so they can enter into their environment. But these RTP sessions are basically not a secure as they need to be either week username and passwords or they are vulnerable and haven't actually been passed. They're taking advantage of those they're entering and there and then once they have that initial access into the network, they spread their payload all throughout the environment and hold all those the those devices hostage for a specific ransom. Now, if you don't have the, you know, particular backup strategy to be able to get that ransom we're out of there and get your your information back on those machines again. Sometimes you actually may be forced to pay that ransom. Not that I'm recommending that you sort of do so, but you see, or organizations are decided to go ahead and pay that ransom. And the more they do that, the more the adversary is gonna say, Hey, I'm coming back, and I know I'm gonna be able to get more and more. >> Yeah, because they don't usually fix the problem or they come back in and it's like a bank. Open bank blank check for them. They come in and keep on hitting >> Yeah >> same target over and over again. We've seen that at hospitals. We've seen it kind of the the more anemic I t department where they don't have the full guard capabilities there. >> Yeah, and I would have gone was really becoming a big issue, you know? And I'll, uh, ask you a question here, John. I mean, what what does Microsoft s A N D. H s have in common for this last quarter? >> Um, Robin Hood? >> Yeah. That attacks a good guess. Way have in common is the fact that each one of them urged the public to patch a new vulnerability that was just released on the RTP sessions called Blue Keep. And the reason why they was so hyped about this, making sure that people get out there and patch because it was were mobile. You didn't really need tohave a user click a link or click and attachment. You know, basically, when you would actually exploit that vulnerability, it could spread like wildfire. And that's what were mobile is a great example of that is with wannacry. A couple years ago, it spread so quickly, so everybody was really focused on making sure that vulnerability actually gets patched. Adding onto that we did a little bit of research on our own and ransom Internet scans, and there's about 800,000 different devices that are vulnerable to that particular ah, new vulnerability that was announced. And, you know, I still think a lot of people haven't actually patched all of that, and that's a real big concern, especially because of the trend that we just talked about Ransomware payload. The threat actors are looking at are Rdp as the initial access into the environment. >> So on blue Keep. That's the one you were talking about, right? So what is the status of that? You said There's a lot of vulnerable is out. There are people patching it, is it Is it being moving down, the down the path in terms of our people on it? What's your take on that? What's the assessment? >> Yeah, so I think some people are starting to patch, but shoot, you know, the scans that we do, there's still a lot of unpacked systems out there, and I would also say we're not seeing what's inside the network. There may be other RTP sessions in the environment inside of an organization's environment, which really means Now, if Ransomware happens to get in there that has that capability than to be able to spread like the of some RTP vulnerability that's gonna be even a lot more difficult to be able to stop that once it's inside a network. I mean, some of the recommendations, obviously, for this one is you want to be able to patch your RTP sessions, you know, for one. Also, if you want to be able to enable network authentication, that's really gonna help us. Well, now I would also say, You know, maybe you want a hard in your user name and passwords, but if you can't do some of this stuff, at least put some mitigating controls in place. Maybe you can isolate some of those particular systems, limit the amount of AH access organizations have or their employees have to that, or maybe even just totally isolated. If it's possible, internal network segmentation is a big part of making sure you can. You're able to mitigate some of these put potential risks, or at least minimize the damage that they may cause. >> Tony G. I want to get your thoughts on your opinion and analysis expert opinion on um, the attack surface area with digital and then ultimately, what companies can do for Let's let's start with the surface area. What's your analysis there? Ah, lot of companies are recognizing. I'll see with Coyote and other digital devices. The surface area is just everywhere, right? So I got on the perimeter days. That's kind of well known. It's out there. What's the current digital surface area threats look like? What's your opinion? >> Sure, Yeah, it's Ah, now it's funny. These days, I say no, Jenna tell you everything that seems to be made as an I P address on it, which means it's actually able to access the Internet. And if they can access the Internet, the bad guys can probably reach out and touch it. And that's really the crux of the problem of these days. So anything that is being created is out on the Internet. And, yeah, like, we all know there's really not a really rigid security process to make sure that that particular device as secure is that secure as it actually needs to be Now. We talked earlier on about You know, I ot as relates to maybe home routers and how you need to be ableto hard in that because you were seeing a lot of io teapot nets that air taking over those home routers and creating these super large I ot botnets on the other side of it. You know, we've seen ah lot of skate of systems now that traditionally were in air gapped environments. Now they're being brought into the traditional network. They're being connected there. So there's an issue there, but one of the ones we haven't actually talked a lot about and we see you're starting to see the adversaries focus on these little bit more as devices in smart homes and smart buildings in this queue to threat landscape report. There was a vulnerability in one of these you motion business management systems. And, you know, we looked at all the different exploits out there, and the adversaries were actually looking at targeting that specific exploit on that. That's smart management building service device. We had about 1% of all of our exploit, uh, hits on that device. Now that might not seem like a lot, but in the grand scheme of things, when we're collecting billions and billions of events, it's a fairly substantial amount. What, now that we're Lee starts a kind of bring a whole another thought process into as a security professional as someone responds double for securing my cyber assets? What if I include in my cyber assets now widen include all the business management systems that my employees, Aaron, for my overall business. Now that that actually might be connected to my internal network, where all of my other cyber assets are. Maybe it actually should be. Maybe should be part of your vulnerability mentioned audibly patch management process. But what about all the devices in your smart home? Now? You know, all these different things are available, and you know what the trend is, John, right? I mean, the actual trend is to work from home. So you have a lot of your remote workers have, ah, great access into the environment. Now there's a great conduit for the obvious areas to be ableto break into some of those smart home devices and maybe that figure out from there there on the employees machine. And that kind of gets him into, you know, the other environment. So I would say, Start looking at maybe you don't wanna have those home devices as part of, ah, what you're responsible for protecting, but you definitely want to make sure your remote users have a hardened access into the environment. They're separated from all of those other smart, smart home devices and educate your employees on that and the user awareness training programs. Talk to them about what's happening out there, how the adversaries air starting to compromise, or at least focus on some of them smart devices in their home environment. >> These entry points are you point out, are just so pervasive. You have work at home totally right. That's a great trend that a lot of companies going to. And this is virtual first common, a world. We build this new new generation of workers. They wanna work anywhere. So no, you gotta think about all that. Those devices that your son or your daughter brought home your husband. Your wife installed a new light bulb with an I peed connection to it fully threaded processor. >> I know it. Gosh, this kind of concern me, it's safer. And what's hot these days is the webcam, right? Let's say you have an animal and you happen to go away. You always want to know what your animals doing, right? So you have these Webcams here. I bet you someone might be placing a webcam that might be near where they actually sit down and work on their computer. Someone compromises that webcam you may be. They can see some of the year's name and password that you're using a log in. Maybe they can see some information that might be sensitive on your computer. You know, it's the The options are endless here. >> Tony G. I want to get your thoughts on how companies protect themselves, because this is the real threat. A ni O t. Doesn't help either. Industrial I ot to just Internet of things, whether it's humans working at home, too, you know, sensors and light bulbs inside other factory floors or whatever means everywhere. Now the surface area is anything with a knife he address in power and connectivity. How do companies protect themselves? What's the playbook? What's coming out of Red hat? What's coming out of Fort Annette? What are you advising? What's the playbook? >> Yeah, you know I am. You know, when I get asked this question a lot, I really I sound like a broken record. Sometimes I try to find so many different ways to spin it. You know, maybe I could actually kind of say it like this, and it's always means the same thing. Work on the fundamentals and John you mentioned earlier from the very beginning. Visibility, visibility, visibility. If you can't understand all the assets that you're protecting within your environment, it's game over. From the beginning, I don't care what other whiz bang product you bring into the environment. If you're not aware of what you're actually protecting, there's just no way that you're gonna be able to understand what threats are happening out your network at a higher level. It's all about situational awareness. I want to make sure if I'm if I'm a C so I want my security operations team to have situational awareness at any given moment, all over the environment, right? So that's one thing. No grabbing that overall sort of visibility. And then once you can understand where all your assets are, what type of information's on those assets, you get a good idea of what your vulnerabilities are. You start monitoring that stuff. You can also start understanding some of different types of jabs. I know it's challenging because you've got everything in the cloud all the way down to the other end point. All these mobile devices. It's not easy, but I think if you focus on that a little bit more, it's gonna go a longer way. And I also mentioned we as humans. When something happens into the environment, we can only act so fast. And I kind of alluded to this earlier on in this interview where we need to make sure that we're leveraging automation, artificial in intelligence to help us be able to determine when threats happened. You know, it's actually be in the environment being able to determine some anomalous activity and taking action. It may not be able to re mediate, but at least it can take some initial action. The security controls can talk to each other, isolate the particular threat and let you fight to the attack, give you more time to figure out what's going on. If you can reduce the amount of time it takes you to identify the threat and isolate it, the better chances that you're gonna have to be able to minimize the overall impact of that particular Reno. >> Tony, just you jogging up a lot of memories from interviews I've had in the past. I've interviewed the four star generals, had an essay, had a cyber command. You get >> a lot of >> military kind of thinkers behind the security practice because there is a keeping eyes on the enemy on the target on the adversary kind of dialogue going on. They all talk about automation and augmenting the human piece of it, which is making sure that you have as much realty. I'm information as possible so you can keep your eyes on the targets and understand, to your point contextual awareness. This seems to be the biggest problem that Caesar's heir focused on. How to eliminate the tasks that take the eyes off the targets and keep the situational winners on on point. Your thoughts on that? >> Yeah, I have to. You know what, son I used to be? Oh, and I still do. And now I do a lot of presentations about situational awareness and being ableto build your you know, your security operations center to get that visibility. And, you know, I always start off with the question of you know, when your C so walks in and says, Hey, I saw something in the news about a specific threat. How are we able to deal with that? 95% of the responses are Well, I have to kind of go back and kind of like, you don't have to actually come dig in and, you know, see, and it takes them a while for the audio. >> So there's a classic. So let me get back to your boss. What? Patch patch? That, um Tony. Chief, Thank you so much for the insight. Great Congressional. The Holy Report. Keep up the good work. Um, quick, Quick story on black hat. What's the vibe in Vegas? Def con is right around the corner after it. Um, you seeing the security industry become much more broader? See, as the industry service area becomes from technical to business impact, you starting to see that the industry change Amazon Web service has had an event cloud security called reinforce. You starting to see a much broader scope to the industry? What's the big news coming out of black at? >> Yeah, you know, it's it's a lot of the same thing that actually kind of changes. There's just so many different vendors that are coming in with different types of security solutions, and that's awesome. That is really good with that, said, though, you know, we talked about the security shortage that we don't have a lot of security professionals with the right skill sets. What ends up happening is you know, these folks that may not have that particular skill, you know, needed. They're being placed in these higher level of security positions, and they're coming to these events and they're overwhelmed because they're all they'll have a saw slight. It's all over a similar message, but slightly different. So how did they determine which one is actually better than the others? So it's, um, I would say from that side, it gets to be a little bit kind of challenging, but at the same time, No, I mean, we continued to advance. I mean, from the, uh, no, from the actual technical controls, solutions perspective, you know, You know, we talked about it. They're going, we're getting better with automation, doing the things that the humans used to do, automating that a little bit more, letting technology do some of that mundane, everyday kind of grind activities that we would as humans would do it, take us a little bit longer. Push that off. Let the actual technology controls deal with that so that you can focus like you had mentioned before on those higher level you know, issues and also the overall sort of strategy on either howto actually not allow the officer to come in or haven't determined once they're in and how quickly will be able to get them out. >> You know, we talked. We have a panel of seashells that we talk to, and we were running a you know, surveys through them through the Cube insights Most see says, we talk Thio after they won't want to talk off the record. I don't want anyone know they work for. They all talked him. They say, Look, I'm bombarded with more and more security solutions. I'm actually trying to reduce the number of suppliers and increase the number of partners, and this is nuanced point. But to your what you're getting at is a tsunami of new things, new threats, new solutions that could be either features or platforms or tools, whatever. But most si SOS wanna build an engineering team. They wanna have full stack developers on site. They wanna have compliance team's investigative teams, situational awareness teams. And they want a partner with with suppliers where they went partners, not just suppliers. So reduce the number suppliers, increase the partners. What's your take on that year? A big partner. A lot of the biggest companies you >> get in that state spring. Yeah. I mean, that's that's actually really our whole strategy. Overall strategy for Ford. Annette is, and that's why we came up with this security fabric. We know that skills are really not as not as prevalent as that they actually need to be. And of course, you know there's not endless amounts of money as well, right? And you want to be able to get these particular security controls to talk to each other, and this is why we built this security fabric. We want to make sure that the controls that we're actually gonna build him, and we have quite a few different types of, you know, security controls that work together to give you the visibility that you're really looking for, and then years Ah, you know, trusted partner that you can actually kind of come to And we can work with you on one identifying the different types of ways the adversaries air moving into the environment and ensuring that we have security controls in place to be able to thwart the threat. Actor playbook. Making sure that we have a defensive playbook that aligns with those actual ttp is in the offensive playbook, and we can actually either detect or ultimately protect against that malicious activity. >> Tony G. Thanks for sharing your insights here on the cube conversation. We'll have to come back to you on some of these follow on conversations. Love to get your thoughts on Observe ability. Visibility on. Get into this. What kind of platforms are needed to go this next generation with cloud security and surface area being so massive? So thanks for spending the time. Appreciate it. >> Thanks a lot, Right. We only have >> a great time in Vegas. This is Cube conversation. I'm John for here in Palo Alto. Tony G with Fortinet in Las Vegas. Thanks for watching

>> live from Las Vegas. It's the queue covering del Technologies. World twenty nineteen, brought to you by Del Technologies and its ecosystem partners. >> Welcome back, everyone to the cubes. Live coverage of Del Technologies World here at the Venetian fifteen thousand attendees. One of the biggest, most important tech conferences all year long. I'm Rebecca, not your host. Along with my co host, stew Minutemen. We're joined by Caitlin Gordon. She is the VP product marketing at Delhi Emcee. Thanks so much for coming back on the cute Kate. I >> know This is so nice. Maybe we'LL have to make it three days in a row. >> I would we would love that. All right, so the last year at this very comforted you lunch power, Max, what's Tet Walker viewers through Sort of. The new capability is the latest and greatest. What's going on with power Max this year? >> Yeah, My favorite thing to talk about his power, Max. So we couldn't miss that today. Yeah, So a couple of updates in the Power Mac's front couple on the software side and then on more on the hardware side as well. S o from ah software side. We've got a couple pieces, which is a lot of our customers, really starting with the largest of our customers, are looking to add more automation into their data centers, and storage is no exception. And how do I automate some of those storage work clothes? Teo, make things run more seamlessly, get into more of a cloud operating model. So we had a couple of announcements on that front. We have a new V R. Oh, plug in, um, to automate work clothes through the r o A. CZ. Well, as ants will play books coming this summer, a couple important automation hand spins and obviously a lot more to come there in the future. The other one in a similar vein, is that containers, right. We've seen the increase adoption of container. So, um, and that the container is being used in production applications means that external storage is actually become a reality in that world, and the support for a C. S. I plug in on power Max, is something that we're seeing more interest from. So we have announced that's coming this summer as well. >> So, Caitlyn, I remember a year ago when Power Mask got announced. I heard things like intelligence and automation. And I went to add non, you know who's been working on this kind of technology for decades? Is that non how we've been talking about this for decades? Tell me why it's different and he lit up like I hadn't seen him in awhile, told me, What's going on for I want you to connect now a year later is what's this mean for customers? What does that automation? You know, an intelligence mean, is there a certain KP eyes or hero metrics who have is two customers using this today that they couldn't have done? And with you, no last generation intelligence storage? >> Yeah. Hey, think about it. It's really about moving to this concept of the Autonomous Data center. And how does this become an autonomous storage system? So both the intelligence within the system that we talked about last year and the decisions that the system is making itself every single day all by itself, that's that has really changed. And it's a completely new evolution of its making billions of decisions a day for customers so they don't have to do that means you're gonna have fewer people managing storage and they can invest in other things. Then when you move that up the stack, some of that the bureau, the answer will play books really enables you to then automate more of the work flows within that so again gets you more into that operating model, and you can automate not just the storage infrastructure, but then get to this autonomous data center >> So way talk to Travis briefly about Dev ops and you're mentioning answerable playbooks. You know, for years we've been talking to customers and saying, Okay, we we need to get two more agile environments, you know, Dev ops there, but enterprise storage specifically, there's a little bit slowed up, so it sounds like we're starting to get to greater adoption. What? What, what what got us over that you know, Hurdle, and where our customers with it today? >> Yeah, and I think it's really the maturity of our largest global customers that have gotten to a place where, for the workloads that will continue to remain on these thes on from infrastructure on our purpose built storage on our high end arrays, they need to run that as efficiently as possible. Um, and a lot of the work we've done to build in a. I does part of that, but really, ultimately they're looking at in there. Three. Terek protector. How do they run things more smoothly? Um, and it's really our customers that have brought that us is a requirement, and we've been able to to support that. >> So how do you work with customers? Mean innovation is, of course, an underlying theme of this of this conference. Talk about how you collaborate with customers to to solve their problems and how you help them. Think ahead what their future needs are. >> Yeah, and certainly Travis, I myself, might our teams, as well as the engineering team, spend a lot of time with our customers in the briefing centre. A lot of in the field, um, really talking about their challenges and the privilege that we have, especially with something like a Power Max platform, is the customers we have. There are the ones that are constantly pushing the boundaries of what we can do for them today, so they always need the best performance. The best efficiency and what has changed is they also now we need that simplicity. They need that operational simplicity, even on their high resiliency. High performance systems. Um, and we spend a lot of time understanding those requirements on DH, the problems that they're trying to solve and how we can help them get there and that that could be automation that could be containers. But it could also be cloud right, And that's the other piece that we've we made a lot of investments across our portfolio is how do we support that cloud consumption cloud operating model, leveraging public cloud? Um, and and a lot of it really just comes from how do we help our oppressors continue to solve their problems? >> It's a competitive marketplace, and, as you said, customers, they want everything. They want efficiency. They want simplicity. They wanted to not cost them too much money. What what's your unique selling point? How do you message this is This is why our solution is >> that I mean, our overall strategy delancy from a storage perspective is that we're way. We'LL have a single product in each segment with which we've compete and each one will be architected for very specific requirements so that we can meet the combination of a price points and it features and capabilities across all these different perspectives and that each one of our platforms is designed to be industry leading in that category. Which is why we have power Max on the high end, the resiliency, that performance, the availability that you know, banks, hospitals, governments around the world expect. But the same time we have mid range pot for us. We have an entry platform that could be sold for under twenty five thousand dollars, right and has a different set of requirements. We have the unstructured business, which is supporting the data. Aaron. That data explosion in a file data, Um, so the The fact is the matter is this. That is all about having the right actor architecture's so customers can have the data in the right place at the right time with the right service level. Um, and that's why we have this portfolio and within each portfolio that were leading in each one of those categories, That's kind of the bigger perspective we have on it. We do not just have a hammer. Not everything is a nail for us. Um, and that's an important part of how we can partner with with our customers to help themselves. Not one challenge, but all the challenges they have >> killing one of the interesting shifts we saw the show is clouds being talked at more than ever at this show. One of the earlier segments we had on we talked about the cloud enabled infrastructure. So things like power, Max, you know, I asked J. Crone, you know, tell me why this is cloud watching, and he gave me a good answer. What I want to ask you your angle on is when you talk to customers, you know how to storage fit into the overall discussion of their cloud strategy. You know what, some of the key business drivers and you know how how's Del technology? >> And I'm glad you said that because Jay and I have had this cloud washing conversation as well as I think that's the unfortunate thing in the reality in the market in the past, probably ten years is a lot of cloud washing, and where we're really focused today is, and we talked a little about this yesterday as well as they say. There's one piece of the how do we fit into overall Del technologies cloud strategy with the Del Tech Cloud. I'm in the VCF integration. We kind of covered that the other pieces that when we look at cloud enabled infrastructure, we're focused on solving really specific use cases that we hear our customers trying to solve today of connecting that data center into a public cloud. So that could be what we call cloud connected systems. The tearing of data from your own promises, infrastructure into the public cloud. Really, that's more of an archiving. This case, a kind of a tape replacement use case that could be dead, remain cloud tear, cloud tearing cloud pools. All the different pieces we have there could be CLO Data Services, right. Offering storage Data services is in a public cloud. Unity Cloud Edition will be one or the New Delhi emcee. Cloud storage services could be another one or even that cloud data insights piece of it. So it's really about solving that solving real challenges about disaster recovery Analytics in the cloud. How do you do that? In a really impactful way? That's simple and easy for customers. >> Yeah, the other Claude related thing wanted to get your take on is many of solutions. I heard on there is, you know, it's VX rail underneath. It's VX rail underneath. It's VX rail under >> you. Notice that >> I did, and you know a way. We had a number of people. V X ray. Lt's doing great, but, you know, if you talk about cloud and the infrastructure that I have in my data center, you know, we've talked Teo, talk to Dell for years. You know, the new power Max last year is underneath some of those admire. Where does that fit in? Kind of CIA and cloud, you know, infrastructure piece. >> Yeah, in a lot of different places. And for Roddy, for reasons, right? Some of us just the high value workloads you need. The scalability, the resiliency, the performance you need the ability to scale your computing your capacity separately. You want to be able to consolidate not just your applications, but actually all your file and sew something like unity or even power. Max, you can have your block workloads and your file workloads there. So we have a lot of customers looking to use traditional three tier architecture, but leverage that in a true cloud operating model from an automation standpoint, cloud consumption model, but also leveraging public cloud computing, right, leveraging the public cloud and really impactful ways, for example, for disaster recovery, eh? So it's really that combining what people love about our industry leading best of reed storage. Um, with that agility of the public cloud is a combination that we certainly hear a lot from our customers of How can I make the best use of clouds? Everyone walks in and say that club first strategy, but it's really about well, how do you actually think about data first and then how do you have a cloud strategy that supports that? >> So So let's talk about the future. I mean, ahs, You said This is what the customer is thinking about right now, but it's your job to think ahead and make sure that you are giving them solutions that fit their future need. So what are you thinking about the solutions that are available today that were really unimaginable five years ago. I think about ahead to twenty twenty five when there is enough data to fill the Empire State Building thirteen times over. How are you helping companies manage the tsunami of data? >> Yeah, and I think part of that is really about again the operations we talked about. Part of that really just comes back to having the right architecture for that type of workload. So this is where I salon actually well before the data era actually was designed for this specifically. So Iceland, created in the early two thousand's, was designed of one file system from terabytes and two petabytes. A single administrator can manage now up to fifty eight petabytes in a single file system. That's game changing when you think about the scale that we're seeing today. So the reason we went to that capacity isn't certainly just cause we thought we could. It was cause our customers were asking for it. Is these workloads in that data that we're talking about autonomous driving center that are just driving the scale? Ability limits, And they're asking for more and more in the most efficient floor print possible. And if you think about that, especially even in the cloud context, there's a There's a combination of How do you leverage that in the in the data center right? And physics means you can't get it up into the cloud necessarily. Um, but then also, there are use cases. They're like analytics of How do you leverage public cloud computing? But then you have that industry leading scale out now, as on the from the storage side so you can combine that. So you talk about something that we talked about here last year, and now we're talking about it a little bit more as well as our integration with Google Cloud platforms. So a lot of our customers are looking to use G. C p for compute for analytics workloads on DH. It's really almost rent your compute for analytics, but you have to have the right storage platform with the right architecture on the back end of that. So what we've done is fully integrated. Iceland, uh, platform and file system through G C P portal. So you could actually combine that public hug, compute and that file system that can support that type of scale. So it's a really unique combination that can help support not only the scale of that data, but also that some of the unique use cases and work loads that are coming out of that >> So Caitlin lot of products here that that would be talking about. Last thing I want to ask is customer customer conversation you have, you know, is data the center of the challenge and opportunity. They have something else that kind of bubbling up. As you look across the conversations you're having that you could have your audience. >> I think at the center of what I hear from customers, Data's in there, but they don't come in saying its data, right? They'll come in thinking about, you know, just trying to figure out how to use cloud properly there. Think about how Doe I simplify things. How do I, um, operate in a way to meet the service levels with a budget that's definitely not getting bigger? Um, and really be as efficient as possible. And it's not, um, some people are looking to go public. Cloud thinking. It's an easy button are there, but it's it's really about How do we change things? Teo run more efficiently and customers inherently to understand, right that the data is at the center of it, and that's increasingly the most valuable asset in the organization. And then they need to optimize their infrastructure to support that, so it really does come down to what? What can we help them to simplify? Optimize. Secure that so that they can truly unlock that. David Capital. >> Well, thank you so much, Caitlin, for coming back on the Cube. That's thanks for having me. Rebecca Knight for stew Minutemen. There is so much more coming up of the cubes. Live coverage of Del Technologies World in just a little bit.