 Hi, I'm Peter Burrs and welcome to another CUBE Conversation from our awesome studios in beautiful Palo Alto, California. With every CUBE Conversation, we pick a topic, find someone to talk about. The topic today is hybrid cloud. A lot of conversation, AWS introduced outposts. We've got Microsoft Azure talking about centralized as well as distributed cloud offerings. Oracle's doing the same thing. A lot of conversation about hybrid cloud and what it means. And to have that conversation, we've got David Floyer with us. David is the CTO of Wikibon. David, welcome back to theCUBE. Thank you very much, Peter. So David, let's start by saying that there has to be a way of representing different options when we think about hybrid cloud. You've done a lot of research in this domain. How are you representing the continuum, the taxonomy of hybrid cloud for customers? Well, on the slide, it shows that there are essentially five different multiple clouds or hybrid clouds. From left to right, it's multi-cloud. And at the bottom of the slide, it says that this is essentially a set of clouds with an integrated network. And then the next is loosely coupled hybrid cloud. And that adds in the data plane where we look after storage and data protection, data management, et cetera. The middle one is tightly coupled hybrid cloud. And that's where the control plane is now tightly integrated along with everything else. And then the next one is true distributed hybrid cloud. And those are the ones that you were talking about. Those are the AWS Outposts, the Azure Stack at the Oracle Cloud Customer type environments. Also, you could put IBM, some of IBM's recent announcements into that as well. And then last, but not least, and certainly the most, one of the most interesting and different is the autonomous standalone clouds that are going to be at the edge. And they have to be autonomous because they can't guarantee network availability to them. So five classes of cloud, each distinguished by the degree to which they share different types of resources, including state, integration, automation, and the degree to which the application is going to be common across each of these cloud types. Have I got that right? Yep, absolutely. All right, so obviously this is, while this is theoretical, in a sense that we're trying to create some way, some understanding about how to represent these things, it's based on some practical observations about where we are in the industry. Let's start talking about multicloud. Who do you place into that bucket of multicloud hybrid cloud styles? So if we talk first of all about the clouds themselves, there would be clouds from AWS or Azure or IBM or Google. Those are the clouds that you start with, you might have one on premise, but the connection between them is just on a network basis. So the people who are doing that would be clearly Cisco is one of the leading people in that area where they already have a lot of enterprise equipment and experience of dealing with clouds across the whole of the area. And they would be the people that are going to be a foremost vendor in connecting those different clouds together on a network plane. Okay, so let's move to the right and talk about the loosely coupled hybrid clouds. Now here we're having more than common network, we're having a common data plane, which really boils down to a common set of data services that are rendered commonly across the different cloud instances. Who's there? So to do that, you've got to be able to have your data services actually on each of the clouds. So you have to have it in software on AWS or Azure or IBM or whatever it is. And two of the people that's probably leading the charge in that area are IBM themselves. They've gone completely software with all of their spectrum line of software in that area. And pure, pure storage have been very aggressive again in putting things up so that they can be reflected in each of the clouds. And there's other vendors that are coming in from a data protection standpoint, data security standpoint, they may not have the full set of services, but they are looking at how they can apply their services across multiple cloud instances. And there's a lot of vendors there, people like Veeam or Rubrik or Cohesity or Dell, et cetera. Okay, so let's move to the right. Now we've moved from loosely coupled to tightly coupled hybrid clouds where we're starting to share a common automation framework, more control, sharing control data so that we can start to understand the state of applications in multiple different locations. Who's leading there? So some of the leaders in this area are some of the traditional ones like IBM, for example. IBM Cisplex, which came out what 20 years ago. More now. Right. That is where you have time and state being shared across a whole number of different instances or nodes within that Cisplex. Yeah, let's talk about that specifically. So we're talking about a global shared memory notion. More than just a namespace, but actually a control plane that has global insight into where resources are, has names for them, and they may be multiple namespaces, but is bringing a common set of controls to that global set of resources. And time is obviously a key aspect of state. What's got to be synchronized. Yes, yes, right. Okay, so if we move to the right to true distributed hybrid cloud, in the tightly coupled, we have a common control plane, but not necessarily common software, common code at the compile level. We're still utilizing distribution formats, maybe specific, et cetera. But now in a true hybrid, or true distributed hybrid cloud, it's common, common. Yes. Who's there? So yes, it's common code. It can run on any node without having to be recompiled or retested. You know it's going to work. And the people in there are the people that we were talking about earlier. It's people like AWS without posts, Microsoft with Azure Stack, cloud customer from Oracle, three large vendors who are using this to use a cloud-first type model in which they can grow the central cloud as quickly as possible, add things to it, and push that down into the cloud customer or the outposts or the stacks. But to be clear, we're not talking about a common cloud experience. We're talking about absolute common cloud service all the way down to the executables so that the same software can run wherever it needs to run. And finally, let's move one step further to the right. This is the autonomous standalone clouds. Yes, this is at the edge. And this is the most different of all of these. It has to be autonomous. If you think about mobile vehicles or planes or even think about a factory or a nuclear power plant, you have to be able to run that assuming that the network is not going to get through. It's on the edge, so it's the most vulnerable to network. So it has to be autonomous, therefore it has to be able to run by itself. And that sort of cloud is mainly concerned with the state, the state of that edge, all of the devices in that edge, or the windmills in that edge, or the factory robotics in that edge, or in military terms, the automated units in that edge, or the drones, whatever it is, you're concerned about the state of that. But specifically sustaining local control of state. Correct. Against a common understanding of how these things interact with each other. So it brings almost a network real-time flavor to it. It is real-time. It has to be real-time, so it's a shared state across, for example, across a city in terms of the traffic lights. You would see multiple of these small clouds in different parts of a large city, for example, which need to communicate with each other. So you have devices which have the inference code running on them and they're dealing with the device onto which it's attached. And then you have connecting all of those devices together to make this overall state system representation of state. Okay, so we've got five classes of hybrid cloud. How is a CIO going to use this taxonomy to make better decisions? So clearly, by making this decision, what we're doing from a taxonomy point of view is making each one individual and different from the others. There's no sharing between them. That means that from a description point of view, we can describe the whole of this industry. We can say how much is going on in each one, who are winners and losers. So that's how we use it. We'll use this to size different classifications, talk about weird, describe competition and all that stuff. But if I'm a CIO, do I think, oh, I got a business problem that's associated with applications and various levels of common data sharing, control sharing, et cetera? Do I use this to help me choose the specific architecture that I need? So the best way that I think that CIOs are going to use this is say, where am I aiming to be? What is most important to me in my business? If it is the edge, then how am I going to go through these? Because I'm not going to get the edge on day one. How am I going to choose my vendors and my protocols and my standards and my data planes and my control planes such that I can get to that particular end point? So within each one, you would want to look at them individually because that's, you know, you're going to put together a, first of all, a multi-cloud environment, but you should be looking into the future as to how you want to traverse across this and who your major partners and vendors will be, your strategic partners and vendors. And we'll use this as we said. We'll use it specifically to size the market, describe competitive factors, et cetera. All right, David Floyer, thanks very much for being on theCUBE. Thanks very much indeed. Once again, I'm Peter Burris and we have been talking about CUBE conversations related to true hybrid cloud taxonomies, Wikibon research. Thanks very much for watching and until our next CUBE conversation.