(upbeat music) >> Hi, I'm Peter Burris. Welcome another to theCUBE Conversation from beautiful Palo Alto. Here today, we are with Muddu Sudhakar, who's a CEO investor, and a long-time friend of theCUBE. Muddu, welcome to theCUBE. >> Thank you, Peter. Thanks for having me. >> So, one of the things we're going to talk about, there's a lot of things we could talk about, I mean, you've been around you've invested in a number of companies. You've got a great pedigree, a great track record. ServiceNow, and some other companies, I'll let you talk a bit more about that. But, one of the things we want to talk about is some of the big changes that are happening in the way that IT gets delivered within enterprises. The whole notion of IT operations management is on the forefront of everyone's mind. We've been talking about dev ops for a long time. It hasn't been universally adopted, it clearly needs some help; it's working really well in some places, not so well in other places. We're trying to bring that cloud-operating model into the enterprise. What are some of the things, based on your experience, talk a little bit about yourself, and then use that as a level into, what are some of the things that the IT organization, business overall, has to think about as they think about modernizing IT operations management, or ITOM. >> Great topic, it's very lengthy. We can go on for hours on this, right? As we are talking earlier, Peter, so I think operations IT management has been around for what, 20, 30, years? It started with, I guess, at the time of mainframes, to client server. But, as you rightfully said, we are in the age of cloud. How does cloud, AI machine learning, and the SaaS services going to impact ITOM, our IT operation management? I think that's, it's going to evolve, the question is how it's going to evolve. And, the one area that you are always passionate about talking about is cloud infrastructure itself, and the word that you use is called, Plastic infrastructure. The underlying infrastructure is changing so much. We are moving from virtual machines to server-less architectures, to containers. So this whole server-less architecture presents such a new concept, that the ITOM as itself should evolve to something new. I actually, I mean the industry word for this is, called AI operations. AI is just one piece. But how do you take hybrid cloud, how do you take the actual cloud substrate, and evolve IT operation management is such a big topic, on multiple areas, and how it is going to change industry. >> So, let's break it down a little bit. So, you mentioned the term plastic infrastructure. We've written a bunch about that here at Wikibon, and the basic notion of plastic infrastructure is that we can look at three generations of infrastructure, what we call static infrastructure, which might be brick, you add load to it, it might fall apart, but it was bound into the application. And in the world, or the era of elastic infrastructure is really where the cloud started, and the idea that you no longer had to purchase to your peak. That the elastic infrastructure would allow you to peak up, and peak down, but it would snap back into place, it was almost like a rubber brick. But this notion of plastic infrastructure, how do we add new workloads faster, is how do we but do so in a way that we don't have to manually go in and adust the infrastructure. That the infrastructure just responds to the new workloads in a plastic way. And snaps into a new form. Now to, we are going to need to be able to do that. If we're going to add AI, and we're going to add, you know, ML, machine learning, and all these other new application-oriented technologies to this. Can't imagine how we're going to add all that complexity to the application level, if we don't dramatically automate and simplify the operating load. And that's the basis of plastic infrastructure. What do you think? >> No, I completely, I think you kind of touched all the good points, but the areas that I can add on top of what you mentioned, is if you look at the plastic infrastructure, the one area is, so far IT operations management is built around a human being, around a dev ops, and around a IT admin. In the new world, it will be 90 to 80% to be done automated manner. Your trading is algorithmic, your in a self-driving car age, but at the IT operations management is around an IT admin and a dev ops. That got to change. I think cloud guys, the Amazon, Azure, Google, they're going to disrupt this because they have to do this in an automated manner, right? So that means, the plastic infrastructure will be able to run workloads, it should be malleable. It's like the, it should be changes shape and form. And it should be that's where the server-less really comes in. I don't want to pick a computer, and rent it for so many hours, that's still a silly concept, I think this whole virtualization, and virtual machines, is gone to the point of server-less. So, all these things. How do you manage the workloads? How do you manage your apps? To your point, apps have to be mapped downstream. I call it, as service maps. How do you build these dynamic service maps for your application? How do I know which component is failing at what point in time? Alright, asking what I call the root cause analysis. Do you expect a human being to identify that MongoDB, or a SQL server is down, because of this hardware issue? That has to be detected automatic manner, right? At least, a root cause and triage it to the point where a human being can come and say, I agree, or don't agree, able to take. Then, the final thing is, the infrastructure has to be, should be, take actions. Allow it to be at the point where the under, once you detect a problem, the infrastructure should be able to say algorithmically, progammatically to an API. I should be able to impact the change. The problem in chain infrastructure today is very much it's very much driven through scrapes and through admins. Can I do that in a programmatic manner? It hasn't happened yet. >> Then it should be, I mean when you stop and think about it AI for example, using AI as a general umbrella for a lot of different technologies that are based on you know, pattern-recognition, and anomaly detection. And all the other stuff that is associated with AI. But we have pretty good data sources in the infrastructure. We know how these tools operate, they are programmable, so they get, you know, a range of particular behaviors. But there are discernible patterns associated with those behaviors, so you'd think that infrastructure itself would be a great source, to start to building out some of these AI platforms, some of these new modeled, what we call data-first, type of applications. What do you think? >> Absolutely, you nailed it. I think, if you remember my previous company Caspida, which was acquired by Splunk. We did that for security. We created this whole area called user behavior analytics. Right? For security, understand the behavior of the users, understand the behavior of the attackers, actual inside it. Same thing needs to happen. >> But all represented through a device. >> Through a device. >> That had known characteristics. So we weren't saying, we're making big claims necessarily about people. Which have, you know, unbelievably complex, but when you start with, What is a person doing with a device? That set of behaviors is now constrained, which makes it a great source. >> Absolutely. So I think it, like given the sources in the IT operations area, if you were think about, for example, looking at the patterns and the behavior of the application, the storage, I call it like, think of like the four layers. You have apps, you have compute, your network, and storage. There are different patterns and behavior you can do it. You can do anomalies, and you can understand the various workflow of the patterns. But I call it the three P's problem with AI machine learning. The P's are, you actually said it five P's. The three P's, that I usually talk about is, the proactive, the predictive, and prescriptive nature. If I can take this data sources, whether they come from logs, events, alerts, and able to do this for those, I can do planning. I can be able to implement what changes I can do as a workflow and full actions. 'Cause detecting is no good, if I can't take an action. That's where the prescriptiveness comes in. And I think that whole area of IT operations management, what needs to happen is mundane with a human being, will be automated. And then the question comes in is, Do you do this in batch mode, or real-time? >> You want to do it in real-time. But let me get those straight. So the three P's that you mentioned where, proactive, prescriptive? >> And predictive. >> And predictive. So, that's proactive, predictive, and prescriptive. And just, you know, to level it out, I noted that all this is based on patterns. >> Yes. >> That come out of some of these infrastructure technologies. So, as we think about where ITOM is going, you mentioned earlier AI, systems management, AI services management. When we think about kind of some of the next steps, who do you anticipate are going to be kind of at the leading, or leading the charge, as we move forward here. >> I think there'll be a new sheriff in town. May not, or to your point earlier, that when many sheriffs in town in this area. The great opportunity here is when all there is a fundamental change like this happens, there will be new players will win this market. Definitely the cloud guys have the right substrate. The Amazon's, the Azure's, and the Google's of the world. They have the right infrastructure, they are all moving towards the plastic infrastructure. They just have to do more on workload management. They need to do more on the AI operations. >> They have a, absolutely a sense of urgency, and pressing need. >> They have. >> Their business falls down if they don't do this. >> So I think those guys will definitely there. Then all the start-ups, right? I think there are a whole bunch of start-ups, each of them will be doing, from a small niche player, all over to platform players. It's a great opportunity, greenfield opportunity. It's going to open up a whole wonderful, new players will come in. Who will be the next generations' AIOps operations vendors. >> So, I'm going to ask you two questions, then. One, do you think the big boys, the HPE's, the Oracle's, the Cisco's, the IBM's, are going to be able to change their stripes enough, so that they can do both? We're tryin' to keep our stall base and upgrade, enhance it, and try to introduce this new cloud operating model? And we'll talk about the start-ups in a second. What do you think? Are the big boys going to be able to make this transition? >> I think they have to, their hand is already dealt. I call it, the cloud is a runaway train, the cloud today is 30, 40 billion dollars. If you are those mega-vendors, you don't, if you're not making on this, something is wrong with you. Right? I mean, in this day and age, if you're not making money on the cloud, with this, with what we're talking about. So what they do is, how can they, either they, have to offer a cloud services, public or a PERT. If you are not doing that, might as well get into this game of AIOps, so that you are actually making money on the apps, and on the infrastructure. So, all those big, large vendors that you mentioned, about the Cisco's of the world, the Oracle's of the, they have a genuine interest to make this happen. >> Got it. So in many respects, to kind of summarize that point, it's like, look, the cloud experience is being defined elsewhere. It's being defined by Azure. AWS, Google, GCP, and these vendors are going to have to articulate very, very clearly for their customer base, the role that they're going to play. And that could include bringing the cloud experience on premise, when and if, data is required on premise. >> Absolutely, and I actually call this cloud should be the aircraft carriers, right? As a world when it settles, eventually it won't have hundred aircraft carriers. You'll have this three or four large cloud vendors. On top of them, the people who manage the apps and services will be few. You don't need 20 vendors managing your infrastructure. So there'll be a huge consolidation game. The questions is, when that happens, the winners doesn't have to be the like c-Vendors. >> Right. >> The history always show the legacy always loses out. So that's where the start-ups have an opportunity. >> Alright. So let's talk about the start-ups. Are there any particular class of start-ups out there. Is this going to, or are some of the security guys who manage services going to be able to do a better job, because they can make claims about your data? Or some of the guys, some of the companies coming from middleware? Where do you think the start-up kind of epicenter is going to be as we see new companies introduced in this space? >> That's a good question, I don't have any one particular vendor in mind. But I think that definitely the vendors that will come into play will be people who can do log management better. We already know the IS Splunk's of the world. People who can do events and alerts management. People who can do incident problem change management, right? All those things, if you look at the whole area. And people who can do the whole application management, as earlier you were talking about the workload management. So I think each of these functions, there'll be winners coming in. Eventually all of them will be offered by one single person, as a full-stack solution for the cloud, on the cloud. The key problem that I keep noticing is, most vendors are keep still tied to the old infrastructure, which is mainframe, or physical servers. Nobody is building this thing for the cloud, in the cloud again. So somebody who has the right substrate to build this, as a playbook, will end up winning this game. >> Yeah, it's going to be an interesting period of time. Now, when we stop and think about, I made an assertion earlier, that for us to build more complex applications, which is where everybody talking about, it's essential, in our opinion, that we find ways to simplify and bring more automation to the infrastructure. If we think about servers, storage, network, those type of things, is there a particular part of the infrastructure that you think is going to receive treatment earlier, and therefore is going to kind of lead the way for how, the rest of this stuff. Is storage going to show CPU, and network? Or is network going to step up, because some of the changes that are happening? What do you think? >> That's a very good question. I think, look, the key think the key pain points for most people today, if you look at the way the complex questions are, if there's a problem in the network infrastructure, it's very hard to triage that, so that area has to be automated. I mean, you can't expect a human being to understand why my switch or network is not performing. >> It's just happening too fast >> Why like, why WIFI is not working on sixth floor and seventh floor. It's a very, so network will be one area, it's highly visible. The second will be in the database and storage area. Just because my storage disk is full, I don't want my database to be down. It's such an old pattern behavior, People will catch those things in an automated manner. Right? So storage, network, because. Where you see the higher level items is when an application is not performing well. Is it a performance problem? Or, why this component is tied to what component, right? Is this applicant is built on a load balancer, and a load balancer is talking to, and the database. Building that map of who's-connected-to-who, that's a new graph, algorithms graph to the unit. That doesn't exist today. So I think what'll happen is how do you manage an application, given a problem, and mapping that. That is I think the number one, that will start happening first. Everything else, people will happen over a period of time. But the apps that are visible, where a user and a customer can see the impact, will happen first. >> Yeah, actually we have a prediction here at Wikibon, what we call networks of data. Where the idea that we're going to the next round of network formation is going to be data assets explicitly connecting with each other. And then using that as a way of zoning data assets. And saying, this application requires data from these places, and then all the technology that allows you to either move it in. >> Right. >> Or keep pointers, or whatever else it might be. So this notion, you would agree then. You know, a graph of data is going to drive a lot of the change forward. >> And to actually take you to that, I actually talk about saying it doesn't require a single class of algorithm. I call it an ensemble of machine learning algorithms you need. You need some statistical, some probablistic, some Markovian algortithm, some Bayesian, and mainly graph algorithms. This data has to capture the behaviors and patterns that you want to put in a larger graph, that you should be able to mine on. That doesn't exist today. So everybody is most often, when they talk about like their dynamic thresholding, statistical, that actually is there in idea operation management. The next level of how do you build a graph, like too big to fail, in my opinion fails. What is it relying on, like if I come to Peter's house. How is your house looks like, the area, one-bedroom, you have two kitchens, You know what I'm saying. >> It looks like a network of data right now. >> Exactly, right. (laughing) >> Okay, so, I got one more question for you, Muddu. And that is, you work with us a lot, and some of the crowd chats you do. You're a great research partner for us. As you think about kind of the story that needs to be told to the CxO about some of these changes, how's it different from the story that needs to be told to the DI team leader? I can imagine what some of the differences are, but you're talking to both sides. What would you, what would you're advice and counsel be to companies that are trying to talk to the CEO about this, or the board, what do you think? What would you say to 'em? >> I think you kind of got it yesterday in the crowd chat. I think the key thing that the CIO or CxO or CEO needs to have this is, we used to call it Chief Data Officer, where the data is the key, that delimit was applied for the overall business. That same role needs to happen within the CIO now. How do I use my data to make my IT better? So that, maybe call it a CIO, a CDO for the CIO, is a big role that needs to happen, but the goal of that person and that entity should be is, How can you do, can I run my operation in a light sort manner? I call it IT as a service. People talk about IT and service. But IT as a service to me, is a bigger concept. >> Let me make sure I got this, 'cause this is crucially important point. So in many respects, we should be saying to the CEO, your data is an asset, you have to take steps to appreciate, dramatically and rapidly, appreciate the value of data as an asset, and that requires looking at the CIO with the CDO, data officer, and saying, your job, independent of any technology or any particular set of ITOM processes, your job is to dramatically accelerate how fast we're able to generate data. >> During decisions. >> Value out of our data being able to utilize these technology investments. >> Absolutely. Because that person, once you have the data as addition, what will happen is, you'll still use the existing process, but it gets you the new insight, What can I automate? What can I do more with less people, right? That has to happen, like if I'm a CEO, he should wake up and say, 90% of my things should be able to automate today, right? >> Okay, so let's talk about last question. You've been, you've led a lot of organizations through a lot of change. We're talking about a lot of change within the IT organization, when we talk about these things. What's one bit of advice that you have for that CIO or leader of IT, and help them take their people through the types of changes that we're talking about? >> Make bets. Don't be afraid of making bets, unless you make a bet you're never going to win. So every year, every quarter, make a new bet. Some bets, you are going to fail, some you're going to succeed. Unless you make a bet, you will not innovate. >> Peter: And understand the portfolio, and sustain those bets. And then, when you've lost, don't keep putting money out. >> Exactly, yeah, keep moving on. >> Great. Alright, so, Muddu, thank you very much for being here. >> Peter, always a pleasure. >> Alright, Muddu Sadhakar, investor, CEO, once again, this has been a CUBEConversation, thank you very much for being here. >> Thank you, Peter. >> And we'll talk to you soon. >> Muddu: Thank you always, and John too. (upbeat music)