 Okay, welcome back. This is Silicon Angles theCUBE, our flagship program. We go out to the events extracted from the noise. This is our exclusive coverage of EMC World and we're in Las Vegas covering all the action of three days of wall-to-wall coverage. We're getting down to the end of day two, which is, you know, dozens of interviews per day. We'll probably have over 15 interviews. We hope we don't hit a wall tonight. We're going to keep on going. We're going to go out tonight. We're going to get the scoop. We're going to go where the action is. This is what Silicon Angles and Wikibon do with theCUBE. I'm John Furrier, the founder of Silicon Angles. I'm joined with my co-host. I'm Dave Vellante at wikibon.org and Amitabh Sarvastava is here. He's the president of EMC's ASD Advanced Software Division. And we've been unpacking Viper all day and now we're just, we started bottoms up. We started with the engineering team and now we're right at the top. So Amitabh, welcome to theCUBE. Thank you very much. Not to be here. Yeah, so you were formerly with Microsoft, you know, in the cloud area, responsible for Azure and the like, a billion dollar business now we're told by Microsoft, so that's good. So you're on to your next billion dollar business, I hope. Yes. They're charging for Viper? Very soon. Very soon. So anyway, big, big show for you guys. I mean, really the software defined meme is really the main theme here at the show. So how do you feel? Excellent. It's a, you know, it's a lot of hard work. A lot of people have been working hard. We've been sort of like heads down and it's good to come out and now and talk about this stuff and not only just talk about it, we actually have the product working. So it's really great. Yeah, so, well, what are you talking to customers at the event? What's their reaction been? Do they get it? Do they see how they can apply it to their business? What have they been telling you? It's been very positive. And as you know, not only here, but we have been running what we call an early adopters program. And the early adopters program was that we wanted to basically give certain set of customers and we were focusing on enterprise and service providers. So we included both, you know, enterprise and service providers in the program and we basically gave them early access to the bits. And so then not only they could hear our PowerPoint presentations, but they could actually use this thing in their environment. Of course, not in the production environment, but in their proof of concept environments there. And then give feedback to us so that we can go help and improve the product. And that feedback has been incorporated in doing that. And the same thing we are seeing here more and more is the feedback has been quite positive and we are very pleased with it. And also the, it's a new space that we are entering into and so people are trying to grasp what software defined storage is. You know, what does this do? What is that different? And so there are lots of interesting discussions. It's interesting how you guys are talking about service providers as well as enterprises. There's a lot of overlap, obviously, but aren't the needs of service providers and enterprises somewhat distinct? And how will your vision of Viper and software defined storage meet those distinct needs? Yes. So if you look at Viper, it does two fundamental things, right? First of all, the foundation that we have built is really built, its design point is cloud, okay? And so the foundation that we have built is designed for massive scale, high availability and geo-distribution. That's what the foundation of Viper is. Then the second part that the Viper does is, you know, it decouples the control plane from the data plane. And by decoupling it, now we can go use the control plane, the Viper controller, to really go optimize a lot of those things out, right? That part is common to both the enterprise and to the service providers. But now we can also add data services on top. And by adding data services on top, both the service providers and the enterprise can use it in different ways. So service providers can use it to offer distinct services, which can be distinguished from the public cloud guys. So for example, if I build an object on top of Viper, you know, on top of Viper, you get the object that what Amazon's S3 offers or Azure's Extor offers. But you get something more. You can get object on file, right? So now you can charge premium pricing for that. So this gives the service providers to offer objects to their customers, but offer something more that the other guys can do. On the other hand, the enterprises can use that same facility and really build these cloud services and then use that as a path to move to the cloud. Different use cases, but can be there. But essentially what Viper does is it's really abstracted that out into data services and really managing the control. So it's a monetization strategy for the cloud service providers, a migration strategy for the enterprises who want to get to the cloud. So coming from that world of what we talk about hyperscale all the time, John says you want to know what's happening in the enterprise in five years, look what's happening in hyperscale. And we've done that for years. We've been fascinated by what's going on there. So how did you take the learnings of, actually let me ask you differently. What's the same and what's different from that environment of hyperscale? So the biggest, by hyperscale, you're talking about the public cloud? Yeah, the Amazons and Azure. Or build your own data centers. So the biggest thing that if you look at, and I'm going to oversimplify it, but it is to make a point, that what the public cloud did was it said, look I'm going to go for massive scale and I'm going to optimize the crap out of it. I mean, just going to optimize like, so you can really drop your optics down and you can really increase your capital. So what they did was they said, I'm going to homogenize, homogenize, homogenize. So they simplified the problem down and they pushed the spectrum into that extreme end there. And it's great because for certain workloads it works beautifully and they make the scale work out. The challenge in my opinion, the hard part that comes in is that in the enterprise, and I ran the vendor server business at Microsoft, so in the enterprise the environment is just inherently heterogeneous. No matter how hard you try it because people have to worry about compliance, they have to worry about cost, they have to worry about security, they have to worry about latency. So the environment is inherently heterogeneous. The question was the hard problem is that what Viper addresses is, can you present those same cloud properties? Okay, and include commodity too, because you have heterogeneity, you've got all of these things, you've got a commodity too. But can you provide the same cloud properties over this heterogeneous infrastructure, which is the, and accept the reality of what exists in the enterprise and really provide them as a path? Because public clouds are useful for certain cases, but you're not going to run your NASDAQ trading system on Azure or S3, because that's not the design point that they would optimize for. So okay, so can you provide those cloud properties over the heterogeneity? Obviously you're going to say yes, but what's the price that you have to pay? Is there a trade-off? And I think that's where it depends upon, you cannot offer different values now, and depends upon how much you want to pay for it. But I can give you objects on a very commodity one, but it's not guaranteed to do that. I can give you object on file, but I can give you object with six nines availability. So I can give you different favors and there's a different price for it. And because you can charge different prices for it, you get different capabilities. So that's why I'm saying is that right now, in the public cloud, it's a raise to zero. Yeah, it's a case. Now we've said that in the Q&A as well. It's a raise to zero, right? And so the question is, it's a raise to zero, it's fine, and you can fight that battle, and it's a money to be made there, and there's no question about it, but it's a raise to zero. Unless you have software. Unless you have software. But what you can also do is expand that spectrum out. Still do the raise to zero. Don't go away from that, because there is value in that. But on the other hand, focus on what are the other premium services that you can offer, which can be monetized in a very, very different way. And enterprise presents you that opportunity because they just have harder problems in different use cases. And that's why we were speculating earlier about OpenStack getting a lot of attention is it gives people comfort as a warm blanket, a bridge to cross. Even though it's not yet defined, it's vendor friendly. Not yet defined, no agendas have creeped in there yet. It's very open source like. So that it looks and it's quacking like the kind of duck people want to see. And so that's a hope there. So my question to you is, a quote we heard here at EMC was earlier, was HP, Dell, IBM, systems will vanish like the mainframe Mini in Unix did. The new platforms, Amazon, Azure, VMware, and dot, dot, dot. This was kind of a quote that was thrown out there. So I mean, I'll see you speculating about other guys going away. So basically the big guys who don't transform will go to the track of the mainframe. Azure, Amazon and VMware, these are the new platforms. Do you agree? And what do they need to do? What is Amazon these guys? That means there's only going to be a few guys running clouds. Well, I think for the public clouds that you've got, you're going to have a set of, it's a very cost intensive business. You've got to build these data centers to do it. And there are a few companies that have the deep pockets to run those data centers off. But then you have a massive number of service providers, which have the aspects that are there. And I think that those elements are going to come in where you have to build those capabilities and let the service providers also come in and offer different sets of capabilities. Well, they have scale too. So the cloud guys built in the ground up, obviously a search engine in a bookstore, create a cloud. So they self built that for their own reasons. And then the rest is history. So that's kind of ironic. But then the service providers all have scale. And they might have milked SMS over the top services and all the things that we know. And they all have their own technology. So they are stovepipes. So my question is to tease out is, how far behind are they, or are they behind? And did they get caught off guard? And what can they do to catch up? In other words, they have some muscle, they have scale, the technologies might not be aligned. Yes. I think the biggest thing that a lot of them, they lack is the technology. If you look at Amazon and Azure, besides the infrastructure and things they own, they build the systems, they have IP, they build these systems, they have all the pieces that are there. Google, they have IP. Vertically integrated. Vertically integrated. They have all of these systems. Microsoft has a lot of IP, they also have a server business. So they have a lot of IP. And so yes, the service providers have massive scale. And but that's the part that they have to develop is the IP part to really compete with these pieces there. And this is the part where we can help them. Is because a lot of the technologies that we're doing, like we are partnering with CSC, that we can assist with different pieces of these technologies out, leverage their expertise in the massive infrastructure and scale that they've got to really bring the space. And then the model is to give them a platform where they can sort of choose, I guess three vectors, right? It's performance, it's availability, and it's function. And that's going to determine the price. That's right. That you charge, and that they can charge their internal clients or their external clients. And they can now build, remember, because by building Viper as an open platform, that they can themselves write the capabilities on top too. So we give them a venue to add new unique values or customize their environments for these enterprises. So this is an interesting discussion. Let's talk about open, because you know, you get on Twitter and you get people criticizing, writing blogs, oh, it's closed, it's locked in, and so forth. So let's talk about open. How do you define open? Why is Viper open? Where is Viper open? Where is it not open? Right. So that's why I remember I would use the word open, I did not use the word open source. Yep. Okay. So the part which you look at Viper is there are three places where value can be added to Viper. One part where the value is going to be added to Viper is the number of arrays that it supports. Can you add an adapter to add a new array to under Viper? Okay. We're going to open those APIs up so anyone can add, it's just like a device driver, really. So anybody can write an adapter to put a new array into this thing, it could be the vendor itself, it could be the customer to do that. That's one interface. The second one is the interface for Viper into the different management stacks. So like we have already done VMware, we have already done OpenStack, and so can you use the APIs so that, so we have opened the APIs up and we have done some of the big ones, but anyone can write their interfaces so that you can interface with any orchestrator or any of those things. So that's the second part. The third part which is there is the data services that can be written on top of Viper. Can I let a startup come in? Aqua comes in, Aqua and these guys which are A4, they come in, can they write their capabilities on top of Viper and monetize their efforts into the stuff there? So we have opened all three of them. What we are not opening up is the source code of Viper itself. Right, so the back end secret sauce, you got to pay for that obviously, that's your IP and you're not open sourcing that obviously either, but you are building this platform in part on open source software and you're participating in open source communities. Is that right? That's correct. Like OpenStack for example. Absolutely, and as time goes on, it's too early. Our intent, at least our intent is to move in that direction, and once we understand how exactly the platform is playing out and things like that and if it makes sense to open other aspects up, we'll look at it, but really the goal is, if you want to build Viper as a close product, then you take a very different approach. But if you want to build Viper as a platform, then you have to take a different approach. Then the only way a platform succeeds is if you're able to create a community around it. And that's the aspect that really we are trying to do it, but at the same time, we are going to make money. Well, we were talking earlier about the old days for us, my early days in my career in the late 80s, early 90s, multi-vendor was kicked around as a word. You know, multi-vendor support. Basically, if you had HP and IBM and EMC stuff, they all have to work together. TCP, IPs, Ethernet adapter, not token ring or whatever. It's just basically connectors, mostly speeds and speeds. You remember the days? We went down memory lane with Pat Gelsinger too. But that was driven a lot by the OSI model, which only really kind of worked up until TCP, IP, but everything was standardized. That's right. And good things happened. So we said, okay, that's that. In today's model, in today's world, what is the equivalent multi-vendor? So we say open in choice is still too vague. And to me, I think what we just discussed there was it's open source. Back then, there was no open source. So open source is the new multi-vendor. Because that's what you're saying. Because open source is the equalizer. If it gets traction, like open stack, that could be the... Well, the question comes back is that there are multiple layers in the system. Like, if you look at Viper, there are multiple layers in the system. There can be certain aspects of it that can be made available. You know what I'm saying? And you're exactly right. I mean, you have the multi-vendor. But you don't have to open source Viper. Right. You just have to support open stack. Yes. That's what I'm trying to get to. So you can be open. Yes. And say, hey, it's going to contribute. Contribute to the open stack. Absolutely. So that's your multi-vendor angle. That's right. It's community-driven. So just tie it to what you were saying. So tie it together. I agree with you. So you can make Cinder better, for example. Yes, absolutely. And it's conscious because of the open source. Yeah, it makes Swift more scalable. Yes. Because our interest is in our storage systems and that, you know, where are the aspects there. So, absolutely. I think your point is better. So I see this as a TAM expansion strategy. Yes. So if you look at, and I hear David Goulden all the time and you used to hear Joe Tucci say all the time, well, you know, we have whatever, 30% of the market, but we're not satisfied with that. We want more. And you guys are by far the biggest markets here, but you're not Cisco of storage. Cisco is about two-thirds of the networking market. So potentially this is a way to dramatically expand the TAM. Now, we'll see how others respond. I mean, you know, if some other large company comes out with a big storage platform, are you going to necessarily write to their API? Are they going to write to yours? Maybe not, we'll see how it goes. Yeah, but generally speaking, your objective is real, I'm trying to get to, if I understand it correctly, your objective is to increase the market that you can serve. Absolutely. As opposed to some companies, I mean, Oracle would be an example, they want to take the Oracle stack and put that everywhere. They're not trying to go horizontal. This is a horizontal place. This is a horizontal place. I mean, like, look at the things that we're adding, like with objects and HDFS, we're trying to enter these new spaces which are there where we traditionally have not played. Was there ever a discussion? I mean, you know, you talk about object and HDFS, listen to the Moretz. I mean, a lot of what he's doing is very storage related. I mean, I look at Hadoop, it's a storage system. So, I think the pivotal effort very nicely fits into the stuff because if you look at it, HDFS is almost becoming a de facto standard, right? So all the analytics that they're going to build on top of it, we just work on a wiper. So that's why if we use the standards, and that's the other thing if you look at it, that, you know, like an object, we didn't try to mess around and come up with our own APIs. We just said, let's just go, just implement S3, right? And we just did S3 or Swift. I mean, there are enough already, right? Enough APIs and enough controllers. Yeah, exactly, you don't need more. So we just went ahead and made sure that, you know, we adhere to standards. So like use S3, this one. So wherever we could, the approach was, just use the standard. So for analytics, HDFS is going to be the standard. Let's just go use HDFS. The advantage we get is that now I can take a lot of the stuff which Pivotal is building, and that now that moves over and starts running on Viper. Right, and obviously include things like Atmos and Centera, which are part of your portfolio. Yes, and we also support Atmos APIs, right? And that's the thing that we unify our Centera, Atmos, and all of the different object systems they've got. They've all got their sweet spots, right? And if you look at, and so this way we can shield it off and let the customers use it, and it allows them a strategy on whatever their business needs are, they can keep moving to different platforms over the future without having to do these massive petabyte migrations to do that. So I have another question on just sort of the control function that you guys have taken out. It takes a long time to build a storage management, volume management, data management, it's hard stuff. Did you do that from scratch? Did you borrow that from sort of classic EMC, a little bit of both? The Viper by itself is completely built ground up, okay? We do have a few components that we use from open source, but besides that, the majority of the entire infrastructure is actually being built ground up. We already had, you know, it's a very, so you remember we have the suites, the storage resource management suites and the SAS suites. We have a lot of the management suites there. So the interface we use is we built in the fact that when you connect to, Viper connects to the arrays, it just collects a lot of data. It's like a host pipe that comes up with all the data splitting out. We can connect that pipe to anything that knows how to analyze it, right? We can connect it to VMware and let VCOups do it, or I can take that pipe and connect it to SRM suite, then SRM suite, you know, because the new version, doesn't care where the data is coming from. And that's the strategy that we went out and used, but all the lower level automation. I mean, we showed one of my, in my keynote, we showed that, you know, like right now, if you take the cover point and you put it inside your environment, there are 33 manual steps. This is our product, and this is all over the place. 33 manual steps, I had it all listed on my screen. 33 manual steps. Now with Viper, it's a checkbox. Yeah, and the thing that Dave and I were talking about prior to coming, doing the QPA this year was, you were comparing contrasting some of the business models around open source, especially like Hadoop and normal Linux, et cetera. People made money by service and support. Yes, they did. Mainly because it was complicated. So what's interesting is the most successful companies by our standard are the ones who are going to eliminate the complexities, eliminate the hassles of integration, and that's the job of the companies to provide value. That's inherently in conflict with the business models or the folks who are promoting it. So a little bit of a collision course there. So, you know, the red hat for Hadoop or the red hat of something might not ever happen again. I mean, what's your take on that? Because inherently there's a conflict there. We want simplicity, we want ease of integration. And look, any time you, you know, even when the public cloud came out, you know, in the, yeah, that's when we even started doing the cloud work at Microsoft, right? A lot of people can come back and say, forget even the business model, even inside people there, I'm going to lose my job. Right, because literally we could manage all of these things with few people and nobody has to pay for it. The API. We never had people sitting in data centers, right? If we're all sitting on one forum, it's a simple, it's few people sitting in Redmond managing the entire infrastructure going on all over the world, right? So you could do that. But any time you happen that, whenever the shift happens, I think the job shifts or the, or the value proposition sort of like shifts there. It's like a, you know, flows out to another area. That's right. The DBA or the break fix guy turns into another position. Yes. Data scientist. And so maybe it's not the maintenance part that goes out, but maybe, you know, the value shifts like, hey, should you go write data services? Do you write to integrate with different stacks? Different or, you know, and like a lot of the enterprises you go through, they have so much custom piece which was there. And that was the feedback everyone said. The reason they were, I was using the word open was where point was, you're not going to come and tell me use wiper, but now you got to use this management stack. Right. Whatever my environment is, you got to plug into that. And enterprises have actually, because they have a lot of money and they have a lot of complicated stuff there. They have built their own environment. Now, I would expect the entire industry is going to move in this direction or else they're going to be in trouble. That's right. Okay, so this is the new, I mean, units used to be open. Yes. Right now it's sort of the other end of this spectrum. Yeah, so I would expect that you're going to get a lot of competition here. I mean, it's nothing really out there that's like them. The closest, frankly, is open stack. Right. I mean, I would say, you know, but as I think you would point out, it doesn't have the functionality. And in my opinion is, I think we're going to be okay. The, like the, we're not afraid of competing. You know, I think it'll be good for the storage industry to come up with various options because that's when the, that's when it gets exciting. And then, you know, things evolve much faster. But on the other hand, with open stack, I think we're going to balance it. Because from day one, when we come out, we're going to be using their APIs, make sure it is there. And if the customers have open stack in their environment, we will integrate. Well, I tell you, there's a law in this business that I've noticed is this, the IT business is elastic. You can cut prices, people will, you know, buy more. And so storage has been a constraint. Yes, it is. You know, both spending disk, we talk about flash all the time here, removing those constraints. Oh, it's expensive, it's expensive. It's drives of value. And this, what you're doing is removing the need to just have the container be the bottleneck, you know? And all the money goes into the container. Now you can develop processes and procedures and algorithms around that container so the data can flow freely. That drives value and I think grows the industry. So the more competition, in my view, the better. So. We need more advanced technology so, you know, keep plugging away. Yeah, great stuff. You know, the software models are changing. Dave and I are really bullish on, you know, obviously flash and what that's going to do to just computer science and programming in general. I think, you know, the notion of how people handle memory management is going to be completely gone. As now you're going to have almost unlimited memory store that's going to change how software is going to be written. So we are super excited. Now, are you? No, I'm very excited. I, you know. We're going to remember EMC. Oh, yes, it's just been amazing, amazing thing. It's great to see this all happening and now the next exciting part is going to be the date with GA, which is, you know, in the next few months. Everyone's all smiles at EMC world and then Gould is like, hey, Tucci, where's the delivery? Let's, yeah. No, not only that. We just stood up on the stage and committed. It's happening, you know. Yeah, yeah. All happening in 2013. It can't go back. Well, that happened in New York, right? Yes, we did. Gould said, you, this year. This year was later. I said, well, you know, it's true with EMC, you know. At the time when you asked the question on the spot, the only answer valid is yes. It's a good question. Of course, yeah. What are you saying? No, 16,000 people, no. Okay, thanks for coming on the queue. We really appreciate it. Where are we right back? Wrapping up the day. We're going to try to sneak in some additional guests here towards the end of the day and then we'll wrap it up at the end of the day. Stay with us. Thanks for watching. We'll be right back after this short break.