 In fact, this is Dave Vellante with Stu Miniman, and this is Silicon Angles theCUBE. We're here live at Gillette Stadium. This is the VTUG Winter Warmer. Pujon Kumar is here. He's the CEO of Pernix Data, a company really specialist in flash acceleration and clustered flash, and we're going to get into that. But Pujon, welcome back to theCUBE. Good to see you. Thank you, Dave. So, you were here at Gillette Stadium. The VTUG has expanded, not just VMware anymore, expanding into this multi-hypervisor world. So how does that sort of trend affect you? How are you guys taking advantage of that? Yeah, so for us, again, we are a solution today virtualizing server-side flash, basically creating a new tier, like conceptually similar to how VMware did for compute and memory, we are doing to flash in the server. So today our solution is agnostic to servers, to flash, to storage, but it is definitely VMware specific because of the nature of the market today. But we are also looking into becoming agnostic to hypervisor. So that's where the VTUG fits in. But as of the shipping product today, it is VMware specific. Flash has been an interesting disruption. I mean, I was talking to Tony Asaro earlier, he was on, and he said, Dave, you and I are old enough to remember the old SSD days, and it was a little tiny niche, and I remember they were big giant boxes, and they had battery backups, and thanks to Steve Jobs, I mean, flash just really became this whole new tier of this persistent data store, and for decades we've seen storage function in particular move out of the host into what became the SAN industry, multi-billion dollar, tens of billions of dollars in business. Now you start to see some of that function moving back, and people are starting to rethink the way they're architecting their infrastructure, thinking about the advantages that flash can bring and bringing that resource, that persistent resource closer to the server, closer to the memory, maybe as a memory extension. Can you talk about that just from an architectural standpoint, what you see happening there, what's driving that, and what should people be paying attention to? Yeah, that's definitely an interesting trend. My own background is kind of funny because if you look at what I did in my past, I built Exadata, and Exadata was about taking compute close to storage. We were basically pushing queries into the storage tier and getting 50, 100x performance improvements, and obviously at the end of the day we were leveraging flash technologies under the hood, and now you kind of take the mirror image of that problem with virtualization and server flash, because obviously this was seven, eight years back, so thanks to virtualization kind of taking over the world, and then obviously thanks to VMware and server flash, now it's time to do the reverse, which is can we move data close to where the application is running? Can we really move storage intelligence into the server because that's where you have VM awareness, that's where you have the VM context, that's where you can do things at a per VM level and kind of leave the shared storage to what it is really good at, which is capacity. Bunch of disks could be read fired or whatever it is, data services, you can do your application DR because that's what people like to do, but essentially move the intelligence from this shared storage to the server, and that's exactly what was the genesis of why we started Pernex, we was like okay, there's a bunch of people trying to leverage flash in interesting ways, but how about we leverage flash in the server and leverage it in a way that is totally seamless to existing environments or even new ones that is coming up and do it 100% in software without kind of shipping a box. So when you have the super slow, relatively slow spinning disk, it's fine to push function out, right? And then so now you've got this flash, not super slow anymore, so the performance impacts presumably of being able to not only eliminate the spinning disk, but maybe even bypass disk protocols have huge potential. Is that, am I thinking about that the right way? And can you talk about that a little bit? Yes, definitely. I think there's a bunch of things, right? We just flash as such a fast media, so you do not want to have it at the other end of the wire, away from on the other side of the network and behind a legacy storage controller or whatever it might be. So fundamentally, by putting it close to where the application is running, you basically bypassed, as you described, the disk, the storage protocols, you have bypassed the network and you basically bypassed the old legacy infrastructure. So by putting anything that tries to put this fast media behind an old controller, it's not really taking advantage of the niceness of it. A simple SSD today can do like 20, 25,000 IOPS very easily, especially when put in the server and that's a whole sand sometimes can do 50,000 IOPS. So Stu, there's a big business that's selling old controllers. So somebody's going to get disrupted here. Yeah, absolutely. Or lead the disruption and eat their own business. Yeah, so Pujon, you guys are really creating kind of a performance tier and I have to think that it really kind of varies what application and it's got to really start with the application for the user. Since you guys are now fully GA with your product shipping for revenue, can you tell us about what you're seeing in the customer base, what applications have been kind of the sweet spot for your early deployments? Right, so our customer base today is extremely broad from the lower end SMB customers to the Fortune 500s, to service providers. So we have seen traction everywhere and from a use case perspective, it's been centered around the workloads that have been virtualized the most and that are IO Hungary. So for example, our top workloads today are SQL, Microsoft SQL Server or Oracle. It's been VDI, it's been a SharePoint exchange. These are the top workloads for us today but fundamentally where we are seeing ourselves going is are people looking at this and saying, okay, this is goodness I want for every application in my infrastructure because ultimately every VM out there is doing input IO and essentially you want to accelerate every virtual machine out there. So we have had cases where people start off looking at us and say, oh, I want to put this in my SQL cluster or my VDI cluster and then very soon expand and say, by the way, how about I throw it on every server that I have in my infrastructure? So folks have built new systems that is like a private cloud system for them which includes SQL Server, SharePoint, Exchange, everything and basically thrown for an X on top of it. Okay, so when I think back to early virtualization days, we were getting such greater utilization out of our compute and one of the big challenges is it really broke my storage. I had to, my growth in storage, especially for use cases like desktop virtualization just really grow. So I'm wondering when you look at the total dollars that customers need to spend, if you solve kind of the performance problem, what does that do to the economics of the entire situation? It changes drastically. So I'll give you some examples, right there. There are customers out there who are looking to build a completely new environment, right? And now they can look at their shared storage infrastructure and say, okay, I just need my capacity there. I need to set our disk there. I don't need to go and invest into SaaS disks which are the high-performance expensive disks. I don't need to go and put some kind of a flash on my sand. Instead, I can get that in a much more cost-effective manner, in a scale-out manner, on the server side at a fraction of the cost. So what happened as a result of that in a concrete example is folks got 10 times more performance microsecond latencies at one fourth of the cost in a massive, like a 32-host deployment by using Purnix. So that's what we are seeing and that's the disruption we are causing by moving that performance into the server. So I'm curious, when you do your deployment, do you find that it can delay the upgrade cycle for whatever storage they have, the spinning disk they have, or are you part of kind of an overall refresh? How do you fit into kind of legacy architecture, some of the newer deployments that are going out there? Where does your deployment fit into that? So there are three big buckets that we fall in today. One is somebody has a legacy architecture, they don't have the dollars to upgrade and they have a performance problem. They basically download our software, put in some flash on the server because we give a fully functional 60-day trial to a customer. And they basically solve the storage problem in their legacy infrastructure by basically using us. The other case that we see a lot is where people are designing a new system and this is a green field deployment and that's where they are leveraging us instead of kind of investing in a legacy storage infrastructure and putting us on the server side. So these are the two extremes. The third one is a case where somebody might not have a real pain point, but they download this and they see better latencies and better performance. So we're rocking out here at Gillette Stadium, I'm not sure the background noise is seeping through a little bit. We got these good mics though, it keeps out most of the ambient noise. But so how do you see this whole business shaking out? You know we talked earlier about the legacy controllers, all the large storage companies have made some kind of play into flash. You guys are very much focused, pure play. How do you see this all shaking out? What's going to happen with the sort of traditional storage base? Is it just going to sort of morph into the flash? And ultimately where will your space vector into the whole market? So we are obviously big believers in the server side, right? But we're also big believers in doing things that work for the enterprise. People have billions of dollars invested in their infrastructure, and the reason we have tremendous traction today is because we provide a solution that seamlessly works in their existing infrastructure, and also works as they go and deploy new infrastructure. So ultimately we are about moving storage intelligence into the server. We started off by doing performance today, and we have the only product in the market that do this clustered server side flash solution. Everybody else is like a niche use case. So that's the reason, that's the kind of adoption we have. But ultimately it's about taking more and more into the server. It's about moving the intelligence into the server because that's where it belongs. That's where your virtual machine is running, and that's where we are headed as a company. Well, and it seems like the metadata opportunity is there, right? Today the metadata is locked up in these different devices, and if you can bring that out and put it into a shared resource, that gives huge potential, I would think, for application developers, and it has implications for performance. Is that a fair assertion? Exactly, and it's about taking that, I talk about the storage controller, and moving that storage controller in a software manner, and building this platform that becomes a distributed storage controller on the server side. That is VMwareware, and that's what you become, and that's where you can basically do a lot of interesting things on the server side. What are the factors that have allowed, storage and software has been around forever, right? I mean, go back to the Veritas days, but it was always huge trade-offs. What are the factors, maybe technically and market-wise, that have enabled this whole software-defined theme to come to the forefront? Why now? Now, I think the big catalyst behind this is its virtualization, and the problem that virtualization created from a storage perspective. You know, we are all used to a siloed environment, Exadata is a classic example. This is a siloed environment where you kind of statically partitioned maybe sometimes for my old TP databases and my data warehousing databases, and I use eight servers for this and eight servers for this. That's how people used it, and now it's more about throwing all these applications into this virtual infrastructure, my big computer, which is essentially a cluster of a bunch of x86 machines. So that has led to this problem being created, and thanks to Flash, which is an amazing piece of technology that can really come and solve this problem. These two things coming together, I think has led to all of these interesting things. So there's a real customer demand, and not to mention your mobile devices. All of the infrastructure projects that happen in an enterprise today are also huge driving forces, like my data is increasing, but I need to access my data also. It's not just I need to grow my data. So how do I really balance these initiatives and come up with an architecture that scales as I grow? And that's, these are the catalysts. So it really is a perfect storm. Do you think it's a zero sum game? In other words, to the extent that guys like you succeed, is it taking away from the market, or is the whole market rise, and there's a sort of a more organized exit from the legacy markets into the new markets? How do you see that? I think there's definitely a shift. It's a big shift, and ultimately people have to think in the context of where the world is today. Gone are the days where I got my big honking sands and big refrigerators, and that's what does all the voodoo. It's gone. It's all about software. It's all about flexibility. It's all about leveraging new architecture. So I think definitely there's a transformation going on in the data center. That's the bumper sticker. The voodoo days are gone. Pujon, thanks very much for coming on the queue. It was great to see you again, and good luck with Pernex data. Yeah, thank you, Dave. Thank you, Stu. All right, keep it right there, everybody. Stu Miniman and I will be back right after this word. This is theCUBE, we're live from Gillette Stadium at the VTUG Winter Warmer. We'll be right back.