 Okay, we're back, this is Dave Vellante of Wikibon.org and I'm here with my co-host, David Floyer, also of Wikibon, Wikibon CTO. This is theCUBE. We've been here all day at EMC World. We're going to be here for three days, wall-to-wall coverage, extracting the signal from the noise, bringing you the tech athletes that we can find. And David, we're going to break down Viper. You've had an opportunity to take a look at this announcement. This is our fourth EMC World with theCUBE. You and I have been here many more times, but this wall-to-wall coverage is working pretty well for us. So you've had an opportunity to dig into it. You've been briefed, you've talked to some of the executives, some of the technical folks. So first of all, welcome. Thank you. So tell us about Viper, software-defined storage, the whole software-defined data center. What is Viper? And we'll get into it. Okay, what is Viper? It's a lightweight software layer provided to separate out the control of storage from the storage itself. That's it in a nutshell, and it's being provided for the cloud service community. It's aimed at that community. Okay, so EMC talking about its pressure least, separating the control plane from the data plane sounds very SDN-like, but what does that mean? It sounds sort of geeky. Break that down for people that's here. Control plane, data plane, what the heck are they talking about? What are they talking about? What they are trying to do is separate out the services, storage services, the provisioning, all the other services that needs to be done. Well, let's just break that down. Provisioning, snapshotting, migration, management. Yeah, all of those types of services. They're trying to separate those out. Replication. Well, the application. You know what I'm saying? Replication. Replication. Yes, so all those storage management and services that you would envision as invoking on a storage array using a controller-based capability. Right, there's two objectives they have. One of which is to separate those out. The second objective, they very clearly said, is to allow the capability of going through to that array and preserving some unique qualities about that array. So for example, if you needed a very high availability, very high performance, you might want to put in an extreme IO array there and use, specifically be able to use that for the set of services above it, but using that very high speed flash. Okay, so before we get into that, so we got the control plane, all the storage management functions. What about the data plane? Let's get specific. What are we talking about there? So we're talking about a data plane of any storage array is their objective, of being able to all storage arrays there. So NetApp storage array, Apache storage array. Well, the one's announced is the NetApp third party. In the future, there will be others, probably, even in the future, something like S3 from Amazon, being a storage resource which you can be managed from this array. A target, essentially. Yes, exactly. So essentially. And including, obviously, all of the EMC storage arrays. Okay, and the data plane, we're talking about controlling the physical placement of data at the back end storage device, or? So what you have from that data plane is the ability to abstract virtual servers from all the physical servers that are there. So you could, in theory, for example, break down an Isilon into a number of virtual servers or even expand across multiple arrays a virtual array across multiple arrays in the data plane. Okay, so before we get into who it's targeted at, so why is EMC doing this? Why extract function out of your controller, essentially rendering your hardware a commodity, and then sell software? Isn't that going to marginalize their existing business? Aren't there going to be performance problems? Why are they doing this? So in my view, the key thing they're looking to do is to preserve their relevance to the enterprise marketplace. And the enterprise marketplace is, sorry, the enterprise marketplace is moving, will move slowly to cloud service providers. So you think this is aimed at cloud service providers? Absolutely. They need to own a position in cloud service providers because, as I said today, for every cloud service provider that we lose, we will lose a thousand customers. So they have to compete in this space and provide a service for those cloud service providers for their long term value. They said that today? Okay, so clearly they see the cloud as a disruptive channel. So the appeal to a cloud service provider is what? They have one of everything, right? We know that about cloud service providers. They got EMC, they got NetApp, they got IBM, HP, Hitachi, Dell, you name it, they got it. Okay, so they've got all these assets on the floor. EMC's coming in with a software defined storage approach saying if I get this right, use our control plane to run anybody's storage or our storage. And I'm presuming they're saying if you run our storage it's going to be a little bit better. They didn't say that, but obviously if I was a salesman. You would think so. Well, but I mean, logically they should be able to integrate better. Well, yes, sure. But the more important thing is from the service providers they want to utilize the assets they have, whether it be commodity assets or specific array assets in the best way that they can. And separate out their provisioning of storage to applications and to their own applications from the storage itself. Make that two entirely different things. So that abstraction then allows them to grow as fast as they want to or with whatever storage they is the flavor of the month. That abstraction allows them to do that. So essentially, I mean I think about Amazon. Amazon is essentially an API to a data center. That's correct. So this is sort of an Amazon-like model in that sense that it's an API into a storage platform. Yep, a lot of the storage from all cloud services will be object-based. But the difference is that if you're Amazon you had to do it yourself. If you were Google you had to do it yourself. If you were Facebook you had to do it yourself. You had to create a one-off environment. If you look forward to most of the cloud service providers being much smaller, being around a particular application, a small ISV, they will not have the resources. They have to do it themselves at the moment. But they will not have the resources to do that. They won't want to do that long-term. So what EMC is saying is we'll provide that for you. We'll be the supplier of the platform on which you can build your particular ISV. Let me make sure I understand what you're saying. You're saying that the hyperscale guys got a zillion PhDs running around where they can figure this stuff out for themselves. They can essentially put resources on the problem to save money, whereas the enterprise doesn't have those people resources. They'd rather spend money. That's true, but here we're talking about the cloud service providers. So you're talking about them as being enterprise. So the target customer of the cloud service provider doesn't have as many resources clearly as Amazon and everybody else. So EMC is playing that role to provide those. That platform so they can layer in services on top and they can plug in whatever other, what, management constructs, cloud operations, cloud ops. Everything is APIs. They can plug in a new API for a new storage service or they can plug in a new API for a new type of storage at the back end. So they can integrate their cloud ops, cloud management systems with the EMC platform and control it, okay. Great, so now what do you make of this? Have they given any detail of the function? Where's the function come from? Are they pulling the best out of VMAX and VNX and BRS? Is it all homegrown from scratch? Is it 1.0 and there's no function there? What's your take? My initial take is there's very little that's actually been announced in detail about what is going on. It's also clear to me that their strategy is to take their file-based arrays as the base for this. So it's the Isilon type arrays as the base. And then from that extract a object storage based on those file storage arrays. In that way then they can get something to market quickly as a solution. Obviously the key challenge there is making sure that they can deliver the Isilon type of file storage at a low enough price. That's one of the key challenges for them. But that's, I think, where they will go initially for their own products. But equally they have to provide the APIs for commodity products as well. And the real competition here is the open stack and open compute foundations. Because if that gets to be the platform of choice and they go through only to commodity servers or things like Cinder there are some ability to go through to Isilon, they will lose out in the control of that place. Yeah, so Cinder for those who don't know is the open stack block storage offering, right? It's an open source with contributions initially by SolidFire and Swift is the object platform. And you said this ultimately will be object. So you're talking about an open interface, a restful interface. Well there'll be both of those. There'll be APIs to each of those. So there'll be APIs for example and that mask APIs to a open stack. They will include that as an API. They want to own fundamentally the control plane and let the market decide if you like what storage is best for what application. So they see that storage at the back end is getting commoditized and they're saying great. I've said this for a while now that storage companies have marked up Seagate disk drives, 10X for years. And they've gotten away with it because they say, well, we add value through the controller, which is true. But essentially people are like, okay, but I don't want to, how many times have you heard this? I can go to Fry's and get a terabyte drive for X. Yet I got to pay EMC or IBM or HP or whomever, tell X times. 10X, 15X times that. Okay, and there's a lot of frustration in that and that whole model is collapsing. EMC sees that, so they're saying let's extract the function. But where is that function coming from? You said there's not really that much announced. Where do you think it's coming from? Are they going to take the best of EMC and put it together? Are they going to sort of build it from scratch, a combination? They're going to take the best of, clearly. They need to take the best of isilon in some areas, the best of flash from all of those different things. They need to extract that and put that in as a layer. They will charge for those. It's not an open source, it's not free. This is not free software. They will charge what the market will bear. Plus that 10X market to disk drives, they're going to make it up in software-led infrastructure. Well they won't get away with quite the 10X but they will be competing against the open stacks and they will argue to the service providers, this is a much better service. We're ahead of open stack. We will put the resources in. Now whether they will win in that battle, time will tell but that's their strategy. Well so why would somebody want software, what we call software-led infrastructure of software-defined storage? I was just writing some things up. Faster application integration, better collaboration, potentially expands the TAM in a market that's under pressure. So by going, allowing other people's storage, sort of the, it's sort of a modern version of the, you know, the IBM Sand Volume Controller, right? Connections to the cloud, speed deployment, all those things. But from what you're saying, it's early days. This is a classic sort of pre-announcement. We're showing the direction. You know, we're going to. We're trying to stop open stack in its tracks. So this is a, we're going to freeze the market on some of the guys that have momentum. But also we, I'm speaking for you, you know, we better start appealing to developers. That's one of the things that I'm sure you saw at the open stack conference when you were there with theCUBE and John Furrier. You saw a lot of momentum in the developer community. And it seems to me, David, that this notion of the data center is an API or the storage platform, storage as a platform with an open API. However you define open, open I guess means you can get into it, right? And you can get out of it. Now open could also mean open source, like open stack. Right? This ain't open source. This is not open source though. Okay, that's not EMC's dealio. But nonetheless, in concept, it sounds like the right strategy. Even though it sounds like the execution is a ways off. I think that this strategy is correct for EMC. I think they're early out. They've brought in a great guy to head that. They, the. Yeah, we're going to have him on theCUBE later. I don't know, tomorrow, Wednesday, Amitabh Srivastava. It should be fascinating to. Yeah, and I can ask him all my Colombo questions and get answers to them. But this notion of how open is open is sort of an interesting one. I mean, on a scale of one to 10, open stack being a 10. And let's see, let's pick an example. Let's pick your favorite UNIX box, interestingly enough. Because that means mainframe is more open than your favorite UNIX box probably today. But being a one, how open is this on that one to 10 scale? I, a three or a four. What they're controlling is the services and they're controlling the access to it and they're controlling the APIs. So they've got a big slice of that. What they are allowing those APIs to go through is to this lowest layer. But you know, where would you put the API emphasis on making it as robust and as fast as possible? Which API would you pick? Yeah, you're going to, all right. You're going to put it on the EMC products, right? No. So they will, they will. So okay, so this is essentially, again, EMC transforming its business model from one of middleware embedded into a controller to software led. But from, I mean, storage has never really been open. I mean, I think Sun tried to make open source storage and it flopped. And Oracle stopped it, yeah. Yeah, well, I think Sun stopped it itself because it just didn't make sense at the time. But open stack, obviously, has a lot of opportunity there. I mean, S3's not open. No, it's not. I mean, it's open in the sense that you got an API, but it's not open source, right? So there's, you know, people who have- Oh, S3 could be a target for this particular platform. Yeah, so it's an interesting strategy. I understand there's been some conversation on Twitter about this, oh, there's nothing there. But to me, what's interesting about this is yet again, you're seeing EMC say, all right, we're going to make a bold move. We're going to try to set the direction of the industry before somebody else does it. You're seeing pockets of software defined out there. Grid store, symbolic, guys like that. But as far as amongst the big wheel, so we got to watch EMC, IBM, HP, Oracle, I'll throw Dell in that mix. I guess throw Hitachi in there too. What are those guys doing in software-led infrastructure? Has anybody announced a more comprehensive approach than this, in your opinion, of the whales? No, I mean, the HP are supporting OpenStack very strongly. Yeah, so HP is OpenStack. OpenStack is HP's angle. Yes. So that's, so they're betting the farm on OpenStack for cloud. They're betting on the farm on OpenStack. In that sense, HP is participating with a more open, on a scale of one to 10, where's OpenStack's openness? It's a 10. Oh, it's a, it's a, it's a 11. It's a nine, a nine or a 10. I mean, it's right up there, absolutely. Okay, but functionally, HP, you know, Classic has not announced a comprehensive strategy along these lines. No, not at all. Nor has IBM, certainly Oracle hasn't. HP are certainly a major, Red Hat is another contributor to that. So Red Hat and HP. With the acquisition of Gloucester. Yes, thank you. Right, thank you. Gloucester, right, yes. So HP and Red Hat are major contributors to OpenStack and they, they buy, buy far and away the biggest contributors to that. So that's major, lots of money. So it's interesting to me because John Furrier, when OpenStack came out said, you know, hey, this is a Hail Mary against Amazon. And he was very supportive of, of the idea. And he obviously, we love open source here at SiliconANGLE Wikibon, but he said, look, this is a bit of a long shot. Well, judging from the OpenStack summit, the pass was connected. Yep. They're in the red zone. Yep. Right? And the EMC has responded. EMC clearly has seen that as the environment. But the other thing that Furrier said, you know, a while back was that in many respects, OpenStack is as much a competitor to VMware as it is to Amazon. That's an interesting observation. And I think this move by EMC underscores that. You said it before, you see this as a competitor to OpenStack, you know, going after the cloud service providers where most of the OpenStack deployments are in cloud service provider. Cloud service providers are an increasingly important channel as we see from not only Amazon, but people trying to compete with Amazon. There's just one thing. I mean, if you take platforms in general, it's rare for more than two to be the survivors at the end of the day. In the days of the PC, it was Microsoft and Intel, the duopoly that won and Apple just survived. Today in the mobile, it's Android and iOS. The two people have survived that. It's unlikely that anybody else is going to get into that already at this stage. What's interesting about this is one or two will survive and it's going to be interesting whether EMC is in that mix as well as OpenStack. You don't think there's more room? You don't think that given the fragmentation in storage, that HB... Not in the control plane. IBM... It's the control plane. If you make that and OpenStack doesn't say anything about who provides the storage. But you're saying if EMC is successful with making Viper the de facto control plane. Yeah, they could be a success. You're saying whoever could do that, whether it's EMC or IBM or HP or somebody else, whomever can create that de facto standard, you're saying could have Cisco-like share in the storage market. It consists of like share in the total IT hardware market. They would define both the hardware and the storage. And where would a FusionIO play in something like this? FusionIO is itself betting that people like OpenStack and Open Compute Foundation will be the winners in this and they will provide certain open platforms and software that will be taken up by those platforms as the de facto standards. So this is the future of storage in your opinion. It's hyperscale-like. It's a page out of Amazon and Google and Facebook. If you, again, as Furrier says, if you want to know what's happening in the enterprise five years from now, look what's happening at Amazon, Google and Facebook today. And you'll see it, and this is what they were doing five or six years ago. They can't afford to make each one of those their own. Right, and so that's the opportunity for IT suppliers to bring them in. What does this mean for CIOs? If this is the future of storage, what actions should CIOs be taking as a result of this? Well, initially wait and see. There's no action. Early, early days. Shake your hands, watch. Watch and see what happens. See who's going to be winning in this marketplace. If you have to make a decision, then you've got good products out there. OpenStack is good, but you have to invest a lot. If you find a solution, like, for example, the Icelon or whatever other flavor you have, which provides the APIs through to object and provides you with ability to provision for it, you need to make that decision now, then make it. But don't switch horses. It's too early in the race to switch horses or pick horses. What you're saying is an early glimpse to the future of the storage business and watch this space. All right, David Flore, thanks for helping us break down the Viper announcement. Conceptually, very powerful. Very powerful. In terms of what's there today, not very viable, is what I'm hearing, but zero viable, but very clever move by EMC to essentially put a stake in the ground. Good move, as you say, starts to make people rethink what they're doing with, for instance, OpenStack or other platforms and puts them on that curve. All right, everybody, David, thank you very much for coming on, appreciate the segment. Keep it right there. We'll be back after the keynotes with my co-host, John Furrier, to wrap this up. This is theCUBE, SiliconANGLE's production live from EMC World 2013. We'll see you in a little while.