 Okay, we're back. This is Dave Vellante. We're live, VMworld 2011, and this is the spotlight called Beyond Storage Virtualization. We got some great guests here. I'm here with Brian Dore, who's the CTO of Savvis, cloud service provider, making a lot of noise in this space and actually doing some great stuff for customers. And I'm also here with CMAC Nazari, who's the CTO of HP Storage. Gentlemen, welcome. Thanks for coming inside theCUBE. I think this is your first time. Both of you inside theCUBE, right? It's quite a production, isn't it? And VMworld, we're here. Well, let me start with you, Brian. What do you think of the show so far? I think it's a great show. This is one of the exciting shows to be at. A lot happening, a lot of new product announcements. It's a good combination of exposure to what's happening in the industry and people you need to meet with. Your space is exploding. I want to talk a little bit about the cloud service provider area and how you guys are innovating and really driving a lot of change in IT. And at CMAC, you're now, of course, you've gone through them from CTO 3PAR, and you're now part of a slightly larger company. Yeah, but slightly is an understatement. Yes, a much, much larger organization. So hopefully you can keep that innovation coming, right? And in fact, there's a lot more resources available, so we should be able to accelerate some of the stuff you've been doing in fact. Great. So Brian, let's start with you. So we just talked about storage virtualization and how it changed things. It was interesting. I observed that a lot of our clients, our end user clients, they were doing server virtualization, but they weren't virtualizing storage. Now you talked, we talked off camera, you were saying that when you started your project, this initiative inside of Savas, and I want to talk more about it, you started Greenfield. So you had the luxury of being able to, they had to rip and replace. You architected it from scratch, you know, for a cloud environment. If I can say that. So start by telling us a little bit about the part of Savas that you're involved in, and the area that you're talking about here, and then what you decided to do, and how you did it, and how your customers are taking advantage of it. Okay. So when Savas started with virtualized storage, it was actually before virtualization was cool. And server virtualization, as we know it today, hypervisor based, really was not mainstream in any way. I'm talking about 2004. That's when we first picked up HP 3PAR, and then it was just 3PAR with the expressed idea that we wanted to fix what we saw as we evolved or began our next generation of managed compute and storage offerings, a real problem in delivering a value proposition via storage. It was just too hard to get the economics to make sense when you were buying and allocating disk on a per customer basis, running at highly underutilized levels. So the idea of putting a virtualization step between the physical storage and the allocation to the customer that would give you an opportunity to insert ultimately policy as the how storage was allocated in real time on a per customer basis, really the genesis of thin provisioning is what we launched into back in that timeframe. And the whole idea that a service provider needs to be able to offer services at scale that are giving economies back to the customer, not just passing through charges was our motivator. So thin provisioning was the enabler for that. So that's an efficiency play. Obviously you're able to take more advantage of the capacity that's on the floor, provision it quickly, service customers. Are you able to start, are we at the point now where you can segment and provide quality of service by different customer levels? Can you actually do that today or is that sort of a future? No, you can absolutely do that. And you see what's happening in server virtualization and in storage virtualization in the same key themes that are starting to emerge, they're just being manifest in different ways. The idea that in storage virtualization with this inserted virtualization layer, the idea that certain physical media aren't allocated to a specific customer, the insertion of a decision point that allows through virtualization a quality of service control that might be dictated by type of storage, that a particular customer data piece of data that gets allocated to is very much a part of the design of this generation, this current generation of storage, taking advantage of virtualization. So you heard Steve Herrod this morning talking about the insertion of policy across the infrastructure through the hypervisor and through VMware. Well, the storage vendors are making it possible to insert that same kind of policy that's going to give us further controls over what kind of cost a customer has, what kind of performance they get, what kind of high availability they get, all that's now possible since we're decoupled. Let's see, Mac, when three parts started, you used to use the term, maybe you still do, utility storage, remember? And back at the time, storage service providers, SSPs were the hot trend. And you guys started the company targeting those types of firms, didn't you? Exactly. Talk a little bit about that and how their requirements are different. So one of the facts that was obvious when the company was started, that allocation of storage was quite difficult. It was, there was a lot of thought that has to be given to the configuration, set aside the disk, create the London, and what you ended up with was something that was quite standard. The idea was to be able to insert a level of virtualization that divorces the presentation from the physical manifestation which are the drives, right? So you could create these much more quickly and because you have this virtualization can actually change the quality of service, for instance, from rate five to rate one from five which had to near line. So the entire design process was around making it highly efficient but also be able to actually create the resources much more quickly and adapt to the world that is coming in either statically or dynamically. Statically by the storage administrator deciding I need to move this to something that is to a higher class, lower class, or have the system react depending on what the system and the policy is set. So design goals from the beginning was be able to actually manufacture these loans and essentially storage as transparently as possible to the users. And that's something that I think was attractive to service providers because configuring management was a lot of work and quite static, right? Did that really fit the workflow of the time where services have to come online really quickly and be torn down really quickly? So Brian, we've talked a lot on theCUBE for the last several months and even last year at VMworld I go back and we talked about the sort of schism between the service providers and the traditional IT environments. Now you started with the green field so you had that luxury but you have an IT background. You know the challenges of dealing with legacy infrastructure. There's a gap in our opinion, big gap between the best of the best service providers and where the traditional, typical average if you will IT is. And in some respects that gap may not be closing. I think if anything the service providers are investing more, growing their businesses. I don't hear a lot of service providers saying well we're a flat budget, you know. I hear we can't grow fast enough. So it's, we stand a reason that you're innovating faster. Can you talk about that within your environment? Maybe even give us some specifics around, if you can, you know, any metrics that you look at. Paint a picture of what it's like in your environment growth and what your strategy is there. Yeah, so you have it exactly right. The idea that service providers have an opportunity to continue driving investment at a time I would say driving investment through cycles of feast and famine that might otherwise hit the enterprise is absolutely what we think to be an opportunity. A reason why most enterprises should think about a service provider for some part of their infrastructure. We are absolutely driving innovation and adoption of new technology and not just adopting what's thrown over the wall but actually getting closer and closer to major vendors like HP, technology vendors, working our roadmaps closely aligned with them, sharing with them deeply our problems with using their technology as it is today and working through those issues and getting next generations of technology with levels of influence that an individual enterprise, I mean certainly there are lots of enterprises who have big influence but on par we're going to be able to accomplish a lot in that regard. And most recently an example of that working with HP is this idea of how we deal with federated environments, peer motion as a way to help us through the very difficult challenge of upgrading tens of arrays. I mean we've got some 53 par arrays in production today and when we have to upgrade an array it takes a lot of effort. When you have to retire an array it's even more effort and when you can do things like peer motion or operate in a federated way those kinds of challenges get much more easily solved but it takes the intersection of big volumes of deployed systems and a vendor, a technology partner willing to work things out that makes that kind of thing happen. Yeah, we've done some research on this area. Do you lease your infrastructure? No, we buy it. You buy it, okay, so I mean you can imagine it's even worse for companies who lease and everything's coming off at lease every three years and they've got to roll their arrays. We quantified this and I'd love to get your opinion that it was a typical size customer. It was a mid-size actually relatively large customer that cost them $50,000 to migrate an array. And that blew us away. We did a lot of work and tried to quantify it and said wow, that's a huge problem. So this idea of being able to have a capability of doing some kind of perpetual non-disruptive migration is really what this federated storage or peer motion solved. And imagine doing that in an environment like a service provider where you don't just have one or two, you've got 50 or 80 and you don't just have one corporate set of interfaces, you've got a whole array of customers with their own operating environment, some in your colo environment, some in your managed environment, some in your cloud environment. All of those need to be dealt with as you deal with the array from a retirement perspective or an upgrade perspective. It is a huge challenge and that's why we need continued innovation at the technology level. So CMAC, we're talking about beyond storage virtualization. So talk to me about peer motion, what that is, how'd you guys come up with this idea? How it's different? Give us a, we've got about two minutes left, give us a snapshot as to what you guys are doing there. So peer motion really comes out directly as a result of talking to folks like Savvis, understanding the pain that they have in migraine data and essentially not having the flexibility that they have across arrays. There's lots of flexibility within the array, but across arrays it's just still sort of islands and the communication between them is not seamless and it requires lots of work. So the idea behind peer motion is you have a series of arrays and they typically are on the floor, overloaded, lots of spindles, they're working as fast as they can and you bring up a brand new array, it's essentially not doing anything, right? So think about moving the data the same way, think of the ESX, you add another ESX, the motion kicks in and all of a sudden you have much better balance among your ESXs and much better service that is provided to the VMs. We think of the volumes or the lungs essentially is another entity that lives within the network and should be able to migrate seamlessly without actually disruption to the service, right? From one storage array to the storage array to react to workload changes, to react to the case where you need to retire a particular array because it's no longer in use or in fact you have even a third-party array that you're interested to migrate from the third-party array, right? So those are the kind of ideas that you're pursuing and the first release will allow you to actually migrate the data from the three-party storage to three-party storage, sort of completely transparent to the host. Now you're not solving the problem of migrating heterogeneous assets, right? It's only homogeneous, right, peer-to-peer. The current generation, that's what we do. The advantage is you're not putting a virtualization layer in there. Exactly, exactly, you know. We all know the benefits and drawbacks of doing that. The advantage is it's native, you know, high performance, no overhead. But Brian, do you have, you probably don't because you started from scratch, but do you envision maybe your peers as having that requirement for heterogeneous and CMAC, is that something that you guys can address in the future? We, yes, we aren't completely segregated in one technology today. And so as you take customers from one environment to another, you absolutely need to deal with that cross-platform challenge. So the more we can push into the array, the intelligence to be other array aware, even when that other array isn't necessarily the same technology base, that's very helpful. And what you see happening, because you're exactly right, you don't need a virtualization layer here. You don't want more technology here. You want to push that inside the array and let that policy and technology rest there. So we need that. CMAC, we have only about 30 seconds. Is that something that's technically feasible in the near term? So as alluded earlier, we have a small percentage of market. As we grow the market, we need to actually migrate the data from the other arrays. And for us, it's absolutely a goal to be able to actually migrate the data from other arrays, not necessarily federate, but actually migrate the data into the federated storage that the three public provide. All right, Brian, CMAC, thanks very much for coming on theCUBE, sharing your thoughts. We are gonna take a quick break and appreciate you guys sharing your insights. Thank you. Thank you.