 Live from San Francisco, California, it's theCUBE at VMworld 2014, brought to you by VMware. Cisco, EMC, HP, and Nutanix. Welcome back to San Francisco, everybody. This is Dave Vellante with theCUBE, SiliconANGLE's production of VMworld 2014. We're here at Moscone, Moscone South, right inside the lobby on the right hand side. Stop by and see us. Abu Bakar Diare is here. He's with HP. He's a master technologist with HP. We're going to geek out a little bit, talk about software-defined storage. Yeah, I'm going to start off. I said several years ago, I said HP's got to get back to its roots of Invent. It used to be in the logo, right? This is your wheelhouse. This is what you do. You're part of that innovation team. So we're seeing, we're hearing a lot of marketing around software-defined, everything. So SDX, okay, and certainly plenty on software-defined storage. So how much of that is marketing? How much of that is innovation? So I wonder if you could talk about the innovation piece of that. Well, there's definitely a lot of innovation happening. I mean, I know a lot of people have been talking software-defined storage in general and it's a big buzzword and it's very difficult to understand. But from a technology side, there's actually quite a bit of work that we're doing to actually make that story true. When you think software-defined storage, you're thinking being able to provision and access storage capacity instantly, at least that's the way I think about it. And I'd like to pretty much abstract hypervisor layers. I'd like to be agnostic to those things. I'd like to be agnostic to virtual machines, possibly running on an operating system and serving storage. What I'd like to think about when I hear about software-defined storage is, provision to me capacity of storage pool that I can use now, that also offers you data services that are enterprise class, right? So being able to handle snapshotting, replication and all those enterprise level data services that you're used to. And so when you abstract all these things and you think about software-defined storage, just as being able to provision capacity quickly, rapidly and make it cost effective, that's where a lot of the technology to enable all those things comes from, right? Okay, so from a product standpoint, that's the virtual storage array, VSA. Talk more about the technology behind that. So you just describe what it enables you to do. Essentially, it simplifies provisioning and management. But it also includes a rich stack, which is, I guess that stack has always been software. Exactly. So what's the enabler there? It's just a lot of coding, a lot of testing. What's the technology? Little bit of both. So basically, you look at a VSA and it's pretty much essentially a virtual machine. You can think about it as like an IO controller that you put on any hypervisor that you want. One of the key things for us is making sure that we're hypervisor agnostic, which is why we're enabling KVM support coming up here soon, right? That's one of our technology enhancements. And so now you can run your VSA on any flavor of KVM. Initially, our initial support for KVM is going to be focused on Helion. I know you and Vish were just talking in your previous segment was just talking about Helion and our strategy there. So our first release with KVM will be focused on Helion. But coming down the line, right? We're enabling general support for KVM. And then we also already support VMware. We already support Hyper-V, right? In terms of hypervisor. So that's the first key. Now that our VSA, our controller is hypervisor agnostic, you can pretty much have it anywhere in your infrastructure, right, delivering data services for you, okay? So that's the first piece. And then the second piece after that is the technology that we put into the VSA. Some of the things that we do that nobody does, things like multi-site technology, for example, that allows us to stretch a cluster across locations and make your volumes pretty much look like they're local to any one side, okay? We're the only ones that can really do that today with the VSA effectively. And our customers rave about that technology. So there's a lot of technology investment that goes into the VSA. So that multi-site stretch clustering, I mean, presumably the hard part there is making everything consistent. That's correct. You're writing to multiple places. Could you talk about that a little bit more? Why is that so hard? Yeah, so the reason that's difficult is because one, you have a network in between and so you have to deal with latencies, right? And then two, you have to make sure that the data is coherent between both sides. So that when you try to access it, at the time you're trying to access it, you're getting the latest updated copy of that data. And so if you're familiar with the VSA and our architecture, what we do is we use network rate and we're able to pretty much spread your blocks across multiple locations. And that's how we do, that's pretty much the technology that's at the basis of being able to enable multi-site, okay? So we pretty much stretch your volume across two sites and we have replicated copies of that data on both sides, okay? And so as you're trying to access this data, you always have high availability and the data is always there for you because of the underlying technology, our network rate technology. And I'm doing that at synchronous distance, obviously, right? That's correct. And it's all done through software. That's correct, it's synchronous distances. Okay. Campus side. Talk about how Flash plays into this whole thing. So we currently support adaptable optimization and this is where we're using Flash more as a tiering, as a tier of capacity. And the interesting thing about our tiering technology is that we don't, like a lot of implementation that you've seen out there, we don't learn the customer data over a period of time and then make adjustments, okay? Our technology is able to adaptively, pretty much on the fly, okay? Look at the customer workload, look at the demand and make sure that these hot blocks are dynamically moved into the Flash capacity. And so that's the way we're using Flash today and we're having a lot of discussions because of our implementation. We're having a lot of discussion and engineering on how we can easily adapt this implementation to maybe using our Flash more as a cache, right? Instead of just a storage tier. But the cool thing about this implementation of being a storage tier that's different than others is the Flash capacity actually is also storage. It's not just a cache, right? That most implementation will say this cannot be used for IO, cannot be used for data. Actually, can you use it for IO but not for data? Well, we actually let you use that storage, that tier, that Flash tier as storage capacity. Okay, so help me understand that. So if I'm using it as a tier, am I manually allocating data to that tier? Or is it all dynamic? Yeah, that's the beauty of it. So we do provide recommendations to customers in terms of what percentage of Flash they should have based on the capacity they're addressing. Typically 10 to 15% is where you want to be at. But basically what happens then is as your data is happening on the storage node, okay? And every storage node will have its tiers. So you could have two or three tiers, right? It depends on you, because we support multi-tiers. So imagine you're in an environment where you have two tiers, a Flash tier and a SAS tier. And each one of the nodes in your environment will have the same configuration. And so what happens is as the data comes into that node, that node can figure out automatically which blocks are the hot blocks and will migrate those to the Flash tier so that you can get better performance on those blocks. And as those blocks start cooling down and other blocks are heating up, we just swap access to Flash, Flash storage. Now I'm thinking about sort of the technology behind this. I wonder if you could give us a little history here because, like I say, a lot of marketing around software defined. You guys have created this product well before the marketing hit. Where did it come from? Was it VSA? Yeah, was it an effort to just sort of lower costs for mid-tier and smaller customers? Was it sort of observing technology trends, a little bit of both? I wonder if you could talk about that. So this is funny. I'll kind of tell you a funny story. And this is an inside joke. But as you know, technology started with the appliance model, right? So we had this appliance model that would go out there and would sell our technology embedded in it. And one of the inside stories that we tell a lot, and it's probably the first time people are going to hear about this live, is that you have a whole bunch of marketing guys going on the road and demoing this stuff and wondering, it would be cool if I could demo this on my laptop. Because VSA is really based on Linux, underlying VSA runs a Linux operating system. And we add all our software on it. And so next thing you need to take that back to the lab and then a couple of people in the lab just starts a weekend project. And now they can run this demo that they want to take the customer on the laptop. And then this goes on until the customer asks them one day, hey, this is what we want. We want to be able to do that. Put that on anything that we want. And it's kind of like the unofficial birth of VSA, so to speak, right? So, wow, that is a cool story. It's a cool story. Okay, so then somebody said, wow, I can make a product out of this. Exactly, and then you realize you can make a product out of this and you can put it on any server to say X86 base. Hey, sounds like a good story, yeah, so. So it's interesting, so as a technologist, you must, I mean, obviously you see what's going on at three-par inside of HP and they've got an ASIC and it's great. Of course, CMC Mac will tell you, we don't want to build an ASIC. We do it because it affords us advantages. So do you see, and so it's clear at some point, right? A lot of those functions are going to go into software, but do you see this as the future of storage? Software defined? Yeah. Well, I mean, like Vish mentioned in your segment, right? We're looking at the VSA as this software defined play for cost optimized environment, right? And then there's other things that you want to look at in terms of customer needs for other environments. But the VSA is definitely right now positioned within HP as our cost optimized software defined story. We're putting a lot of resources behind it to try to make it easier to use, easier to deploy, and easier to consume, right? And so a lot of the work, like I mentioned to you in the beginning of our segment here, is pretty much going into abstracting the presence of the VSA itself. No, ideally we wouldn't even want you to think about the VSA there. We would want you to think about storage pool that you can dynamically provision and tear down as you need them. And so that's really where we're going. But Bukkar, what do you make of this notion? Because what you just described leads me to VVOLS, right, which is ultimately I want to pull a storage that I can have connect to sort of a VM sort of application view. That's great, that's great. So what do you make of that trend? Is that good news for you guys? Yeah, it's great news. Especially when a VSA environment for us, it's great news because one thing that block storage have been struggling with with VMware in general is that VMware awareness, okay? Being able to look at a container and say, this is a virtual machine. File could do that, NAS could do that before because all these VMs sit on the NAS file share as files pretty much, right? All these VMs. But with block storage, we see blocks. We understand, you know, LVAs and that type of thing, right? And so being able to see where VM begins and end is a little bit more difficult. And then being able to associate data services based on those boundaries is a little more difficult when you're at a LAN level. And so with VVOLS, we now finally have that capability with block storage where we can kind of look at a VVOL and a VVM's definition and all the metadata that goes with it. And we know what a virtual machine now is, okay? And then we can assign all bunch of data services to that virtual machine. We can enable QOS, we can enable stretch replication, we can enable snapshotting. I mean, there's a lot of things that we could do now more effectively in the array at a virtual machine level we wouldn't be able to do before. By bubbling up those services and making them available to the VVOL environment. And then, yeah, that's great. Does that level the playing field then with sort of NFS? Well, not only levels the playing field, but now I think it kind of really leaves it behind because if you look at some of the technologies that we have that they don't have, we really kind of finally be able to showcase the value of these technologies. Specifically, you're talking about the data services? Yeah, the data services, exactly, yeah. All right, we're running out of time of the car, but so let's talk about briefly OpenStack and Helion. What the play is there. So Helion is, as you know, this big OpenStack orchestration cloud play that HP's got initiative that we have. And as part of that, the VSA, the KVM VSA that I mentioned that we've enabled specifically for HP Helion today is at the core of what Helion's going to deliver for the storage stack, right? For the storage stack. So any customer that consumes Helion and deploys it, we'll have store virtual VSA and our, pretty much our API stack has been enabled with this OpenStack sender driver for all these ice-cazi functionality. So you want to deploy storage, you want to quickly provision virtual machines, meaning VSAs, all of the things will be enabled through Helion. And the VSA will be at the core of that. All right, and that's pretty fundamental to HP's cloud strategy. Exactly. All right, Abu Bakr of the R.A., thanks very much for coming on theCUBE. It's a pleasure having you. Thank you, Dave. All right, keep it right there, everybody. We'll be right back with our next guest. We're live from VMworld 2014. This is theCUBE.