 From our studios in the heart of Silicon Valley, Palo Alto, California, this is a CUBE Conversation. Hi, and welcome to this special CUBE Conversation. I'm Stu Miniman coming from our Boston area studio and joining me for this special discussion of hybrid cloud monetization innovation white space. I have James Kobielus, who is our lead analysts out of the DC area and also have David Foyer, who is the author of Wikibon's hybrid cloud topologies coming to us from San Francisco. Gentlemen, thanks so much for joining us. You're welcome. All right, so we've been talking about hybrid cloud quite a bit and the reason we're doing that is if you look here in 2019, public cloud, we think we understand it. It's well-defined, we've talked about it. We can talk about the winners and losers. We can talk about innovation and growth. There's a lot of money there. A couple of years ago, we spent a lot of time looking at private cloud. We put out research, what we call true private cloud. And the reason we called it that is because how do we differentiate it from virtualization, things like the operating model, how we buy it, how we consume it, how we management, all is different than the traditional data center. But when you talk about hybrid cloud, it's interesting. When you think back to the early days, it was, well, I take public, I take private, I put them together and I have hybrid cloud. But is there much more than just selling those pieces? And earlier this year, David Foyer put out the hybrid cloud taxonomy, really put together some of the planes that we should be examining as well as the spectrum of the offerings. So, and today we're gonna talk about some of the opportunities to really up level and where there's opportunity for customers to get a lot of value. And therefore the ecosystem will also be able to get some revenue and monetize that. But David, to kick things off, if you could just give us kind of the primary thesis of that hybrid cloud taxonomy for those that haven't seen it before. Okay, so we've created the taxonomy, which if you imagine it going from left to right, is multi-cloud and then loosely coupled hybrid cloud, tightly coupled hybrid cloud, true distributed hybrid cloud and autonomous standalone clouds. So those are the five labels that we've given to them. And if you think about those five labels, there are four characteristics which increase from left to right. State, for example, the autonomous standalone cloud is all about state. It's about the state of a IoT or a moving car or whatever it is. Integration, the level of integration, automation and hybrid applications. So those are the four levels which change and then covering of those, we have the concept of planes. So we're looking at different types of planes, network planes, data planes, control planes and there are other planes as well. And looking at the functionality that's in those planes which has to increase the further to the right to go. All right, thank you, David. And there's a lot of nuance and complexity and a lot of areas to kind of dig in. But before we go into any of the specifics there, can hybrid cloud be more than just a composite of the various pieces? Where is their innovation? Where is their white space? Where is this going to be more than just say the world of the past where we talked about multi-vendor? Well, I think the way to answer that is to think about what is the most important characteristic of data centers and the characters to computing. And that is what is different now is data. Data is spread everywhere. You have data at the edge. You have data in your own data centers. You have data in the clouds. You have data in SaaS clouds. Data is distributed. If you want to maximize the value of that data, then you need services, not to just pull everything up to one cloud. That's one model. But the key disadvantage of that is it takes time. It takes time and a lot of money to move all of that data up. So a better model is to move your code and services to that data. And to do that, you need a hybrid cloud mechanisms to achieve that. You've got to be able to send that information across the network in some way and you've got to be able to coordinate where and how you send the code and the services. Yeah, well, David, we've been talking for a couple of years now. Data is really at the center of all of this discussion. You talk about cloud. We've talked about big data for years. Want to pull in Jim. So Jim, data is your world. You're talking developers and data scientists and of course the giant wave of AI has data at that center. So you've been looking at and been writing a lot about the role that data and AI play in this hybrid cloud discussion. So up level us a bit to some of the apps in the data as to how this fits into this overall discussion. Well, yes, yes. Well, AI, of course, rides on data and really AI rides on an entire pipeline and workflow of data from sources through to data lakes where the models are built and trained to and then the serving of the built out models into downstream applications which also have their own local data and so forth. So we look at hybrid clouds, they in multiclouds they quite often bring a lot more complexity by their very nature into the whole AI development and operations pipeline. But there's value to be gained from using distributed data for all manner of AI applications. The applications become more powerful because they can leverage more data. You can build more types of models to do more kinds of inferencing and so forth. So when you look now at the multiclouds we have distributed data, distributed models, distributed workloads that are doing everything from natural language processing to face working mission and beyond. That's the power of AI. Now in a hybrid cloud environment you need, like David said, these planes to enable a degree of integration. One of the planes is the data plane and the data plane at its very heart involves moving the data where it needs to be for training the model for doing validation of trained models to make sure they continue to be fit for purpose and so forth. And much more of the training is happening at the edges now because that's where the actual AI is living because that's where the data lives and dies. So you need to make that happen. That data plane needs to be highly versatile. Increasingly needs to be able to work in more of a meshed multicloud environment. And that's sort of the bleeding edge of Istio and all those other things that we're seeing coming into the mainstream of AI and data management in the multicloud. Yeah, great, Jamie. I'm glad you brought edge into the discussion because often it's the battle of public versus private cloud. And that's not really the discussion. It's about my applications. It's about my data. It's where things naturally gonna live. When we've done lots of interviews where you talk about just the natural laws of the cloud. There is latency and the speed of light are still very important things. David, I've spent many years talking with you about these pieces and the role is to what lives in my data center, what lives in the public cloud, what ends up the edge, where code goes, where data goes is a complex thing. But David, I think we're gonna spend most of our time talking about this data plane. Help us peel the onion a little bit on this. One of the services that should be extracted from the individual platforms that we've looked at is data protection. Is that part of the data plane? Is it the main piece of the data plane? How does that and other services fit into this discussion of hybrid cloud? Yes, data protection is obviously a key component. And I think a good way of thinking about it is that there will be, in most cases will be multiple data planes. You will want a data plane, for example, that helps you with high speed record moving or high speed small pieces of data moving around, which is maybe of transient use. The type of characteristics, the network to support it, the protocols and the distributed management of that is gonna be very, very different from, for example, data that you need to get ready for an AI example, where you're training something. So where bandwidth is much more important. So there will be, in my opinion, multiple data planes. There's a big opportunity here for vendors because you need to be able to run your storage services across all the different hybrid versions that you have. You need to be able to run that at the edge. You need to be able to run that in your local private cloud. And you also need to be able to run that in whatever other clouds, public clouds as well. So a key characteristic then is all of this is gonna be code. All of this is about moving those services. And if you think about the types of services that you want in that environment, if you want to write a hybrid application where you are having data in one place and computing in one place and data in somewhere else, you need to know the latency, for example, how far away in time is that data? Should I move the data here or should I move the code to the data? You need operational systems. You need orchestration that will manage that for you. So you need different courses for different courses. You will need different data solutions for different types of application. Different ones for AI, different ones for transactional processing, for example, complex one, like maybe that you have a transactional system and you want to, at the same time, to do a large amount of fraud detection. That's a different type of application and different databases associated with that. If you're at the edge, you're gonna be using far more time series databases, state databases, whereas you'd be using traditional SQL databases for your systems of record. Does that help in explaining? It does quite a bit, David. And yeah, I would say people, please check out wikibon.com. You can see some of David's research. It brings back to mind just the discussion we had on software defined in the infrastructure space, pulling these services away from just hardware and being independent of that as a discussion. David, you've been having for decades and I've been reading some of your stuff from even before I joined the team. So it's great to see kind of what's the same and what's different as we build these out. Jim, the piece I heard from David is, it's a lot of, there's the database and there's all of these services that we have, but at the core of it, when I go back to that data, one of the things we've been grappling as companies is how can I actually monetize that data and get value out of it? It was one of the things that we talked about, big data was supposed to be that bit flip from, oh my God, how do I deal with all that data to, oh boy, I've got all this data and I can do something with it. And it had rather mixed results. So for AI, there's lots of companies out there to help unlock the value of data. It's going to be the rocket ship to drive thing forward. AI is going to drive all this. So it's a complex environment and there's lots of things that companies would have to spend. So are they going to have to spend more money or can they actually make money when it comes to our data? Give us a little bit of an insight as to what you're seeing there. Well, yeah, well, monetizing the data, Stu, where AI and machine learning come into the picture, is that essentially these are statistical models that are derived from the data and provide monetizable value if deployed into working applications. So that's what the whole data science pipelines all about is taking the source data, looking for the predictive variables, building models to incorporate those predictive variables to do, you know, face recognition or natural language processing, et cetera, into models that are trained and deployed into working applications to do inferencing. That's all the monetization train as it were for AI. And so what we're seeing, what Wikibon is seeing is that there has been in the last several years a growing niche in the industry of workflow tools, DevOps tools for the data science pipeline that handle all of those processes-enabled teams of data scientists and data engineers and subject matter experts and so on to work together to really industrialize the process of extracting that value and building those models from the data and deploying into all manner of applications, including deploying these models AI over Kubernetes and, you know, within containers and in serverless environments. This is happening. I mean, we're seeing all over the world in every industry enterprises setting up very much industrialized processes for incorporating data science into the very heart of application development. Jim, that's awesome. So I'll go back, as I've heard Peter Burris talk about this, the thing that differentiates a company, you know, before and after they've gone through that digital transformation is to how they can actually leverage that data. Are you a data-driven company? Is that something that you can take advantage of inside what you're doing? So we've only scratched the surface. Jim and David really appreciate you talking about that. What my call to action for people out there is if you go to wikibon.com, you're going to see the regular ends of research coming out for the team on this. We actually have regular conversations with the community and welcome your input. If you go to crowd, excuse me, crowdchat.net flash action item, you'll find the latest and you'll see if you look on the right side, so some of the previous ones. And of course, our team is, you know, at lots of events, all the big cloud and infrastructure and software shows, check out thecube.net for all of those for Jim, David, I'm Stu. Thanks so much for watching. Please reach out to us with any question. And as always, thank you for watching theCUBE.