 The Cube, an open-stack summit at Lata 2014, is brought to you by Brocade. Say goodbye to the status quo and hello to Brocade. And Red Hat, here are your hosts, John Furrier and Stu Miniman. And we're back. Talking about the software-defined enterprise, what's going on in cloud computing, the segment we're gonna dig into now is an interesting startup out of the valley that sits at that intersection of software-defined environments, what's happening in the open-source community. Joining me for this segment is Doug Faustrum, who is the Senior Director of Product Management from Coho Data. Thanks for joining us. Thank you, Stu. Very exciting to be here. Yeah, so for those that don't know Coho, Andy was one of the founders of Zen. And you guys have a software-defined storage solution that also has software-defined networking involved in how everything's laid out. So, that's my thumbnail on it. Can you tell us a little bit about yourself and tell our viewers about Coho? Sure, so my name is background is product management. I've been doing product management storage for about 15 years. So, come from the Veritas background, doing block storage, plus the file system, the old Veritas store, and realized a while back that IP storage is the future and Coho had a fantastic store in the way to integrate a software-defined networking and the storage stack. And it's really what's doing a very nice twist and very different and very useful twist from a customer perspective, where they're able to simplify the environment by integrating these two typically very disparate components. Yeah, so unpack that a little bit for us. So, my understanding is that the software-defined networking is really a tool inside your solution that helps with the storage. You guys aren't trying to not just go off of their perch or anything like that. Yeah, so we use the standard open-flow protocol to integrate into today's and Arista switch. Just happened to be what we picked at the time and we're hardware-independent from that perspective. Our intention is not to provide a software-defined networking hardware. It's merely using the functionality within the switch itself. Okay, so why is Coho at the open-stack summit? Well, open-stack is if you go back 15 years, Linux just got started. We saw the same momentum open-stack a couple of years ago and it's an extremely exciting conference because it has the momentum of changing the industry and the way it's being managed and run inside a data center. Coho's architecture fits extremely well into the scale-out, new thinking of doing storage management, where we have an all-big-big storage architecture in the back with scale-linearly SEI boxes and it fits really well into the open-stack mantra, open, compute, open-scale and so forth. Okay, so couldn't expect to see Cinder support from you guys sometime in the future? Absolutely. Cinder and Glance and even exposing all-big APIs. Okay. That'll be further out because we've got folks on the Cinder and the Simplicit of use. I mean, something that unique with Coho is the simple to use, the single access point, the single, very easy to use UI. We want to maintain that as we introduce open-stack support into the product. Right, so Doug, I mean, you guys part of the developer community that's building this, are you committing code to open-stack? We will be. So this is our first foray into open-stack. We're super excited to be here. So come, as VM world rolls around, we'll probably see more announcements of us around open-stack and also us joining the community and so forth to provide code and provide the expertise from the storage and Zen background into the open-stack community. All right, so since you brought up the VM where let's talk, where do you guys sit with kind of virtualization? Some of the discussions we've been having at this event, obviously there's plenty of customers that have VMWare, want VMWare, are going to do VMWare on open-stack. Some in the community are looking at open-stack as a way to make an infrastructure that I can put KVM on it. And then there's others that are just saying, I'm going to use bare metal and containers to be able to do open-stack. Are you guys agnostic to that or where do you sit in that space? That's a great question. We're definitely agnostic. We choose to start with VM and merely because they're dominant market share. Now with open-stack, we wouldn't use KVM into the mix. Come that, we also look at HyperV because that's also a big part of market as well. And also clearly physical access to our storage whether it's NFS protocol or SMB protocol is extremely important to us. And that's probably following a timeline similar to that. All right. So give us the company, can you just give us some of the basics on it? When did you guys come out of stealth or are you guys GA with your product, funding rounds, the ML like? Absolutely, so the company's been about almost three years into making their products, I should say. We came out of stealth about six months ago, so October last year. Funding has been 10 and 25 million in two rounds. And the first GA of the product came in about two months ago now. So right when I joined the company, we went GA with the product and one one coming out this week. So we're moving very rapidly. That's expected by startup, to be honest. And the innovation pace is very, very fast. And that's why I'd open-stack support is also gonna come very fast. All right. What are you hearing from your customers in regards to open-stack? A lot of interest. So a big part of customer base is service providers. And as seen here and some of the biggest sponsors are open-stack, it clearly serves providers as a big key piece of open-stack. And so we're seeing service provider interest, the pay-as-you-grow model fits really well in that environment. And as well, that's the open-est open-stack. And be able to integrate it and as you grow your environment, you can easily grow your storage and compute a network. Okay. That's where you're coming from. So service providers, what's your scalability story then? Can you talk a little bit about how you scale and how big you can scale? Our scale is very linear today. So you can run a simple benchmark just to run as much IOS as you can to our solution. We get about 180,000 IOPS, 4K, random, 80-20, you know, mixed out of one 2-U chassis. And as you keep adding chassis, what we've proven is you can scale. And we've tested in that benchmark up to 10 chassis. So 20 microarrays, what we call microarrays. And that scales linearly. So we get over a million IOPS from that environment. Well-passed in my million IOPS. And it's nice to see the linear scale. Would you see what we do in the background as we scale, this is where the SDN switch comes in. And the software-defined network is that as we're able to reprogram the data path behind the switch and therefore move clients from an overloaded array into a newly brought up system, automatically without any system intervention. So if you look at a curve, you'll see a straight curve, a little dip when we do rebalance and a linear scale up. And we also have a paper on that where people can read the details from ESG and certified external analysts. So the interest is being able to add it in without any intervention. You literally plug in, physically turn on, wait 15 minutes, and that's it. Sorry. Can you talk to us a little bit about how management of your environment is done? Are you still considered a traditional storage box that a storage team does? How does the networking team play in? And with OpenStack, usually looking at more kind of orchestration, automation, how does your solution compare to kind of a traditional storage array that people are familiar with? Right. So we try to think things very, very simple. So our mantra with our customers is to keep things in a no-touch environment. So we have a extremely simple to use UI where you just with a few clicks you can see what's most relevant, what is maximum performance, what's using the most capacity, and so forth. From a comparison to, let's say, a traditional vendor like EMC or Touchy, where you're rolling a box, you spend some time configuring it, you spend some time creating groups and volume groups, we don't have that at all. We literally plug it in. First time set up about an hour from literally power on to when you're done. And then all you do is keep plugging the race in. And everything is taking care of in the back end. And we have no user intervention required at all. So it's a very different experience as you move on. And as we move new hardware architectures into the box, as we move to the next version, it's all fully supported to run different versions. So the migration doesn't really happen. We don't need it. You buy another hardware, you plug it in, and our system will automatically see when you need to move data to it. If you need more performance from your box, we'll automatically relay that and move the data behind the scenes. So again, no user intervention at all. So we're taking a more of a cloud-like mantra. So when you go to Amazon, you just say, I want this many CPUs and this much storage. And that's exactly what we're thinking. And this is where the integration at the OpenStack is so exciting to us. Because we obviously don't do the commute. We don't convert storage box. We do storage only. Clearly, we use network into simplified environment. But we also want to benefit from that deployment of a customer. They say, I just want this much storage and this much compute. And we simply just follow those rules. So, Doug, we've seen huge growth in the converged infrastructure market. Trying to simplify that environment. Do you think you guys do enough if you're only focused on the storage piece of it? If you look at storage markets, it's very, very big. We believe there's plenty of room for us to grow. At some point in the future, perhaps, there'll be room to grow in different directions. But for now, we're going to be focused on storage and really revolutionizing that together with the next evolution of the data center. We're just not the old create-a-lun and map-a-fast system to it. Just to create everything as a pool and share storage as needed. Doug, I definitely agree. There's huge opportunity in the storage market. Flash has really kind of shaken everything up. Everything from server-based solutions to all flash arrays. I haven't heard you talk much about flash. What do you guys think about flash as the architecture and how does that fit into your environment? So, flash to us is extremely exciting. We were able to use flash in a way that we couldn't imagine using storage in the old days. Today when we have flash, we're able to metric, for example, every IO and trace every IO coming into the box without any over at all till they're all capacity out of the box. So, we use flash as a tier zero storage area today where everything ends at the end of the box, ends up on flash. And as data becomes less relevant for the user, we move it off to a hard drive environment. Today we are hybrid array. Doesn't mean we'll say hybrid forever, but it's definitely where we're at today. We use flash extremely fast and low latency IOs. And that enables us to do things we couldn't imagine possible with storage five years ago. It's really, really interesting. What's most interesting part is that you can saturate a single 10 gigi interface with a single flash card. And that's something you did not see three years ago. And we think that will change even going forward in the future. So, finding a system that provides the right balance between network, CPU and flash, it's going to be critical for scale. Yeah, absolutely. We've had a big mismatch and server virtualization came in, it caused a lot of challenges with storage and networking. Now storage with flash is causing challenges with networking and we're trying to do that. You know, let's talk about the economics of the solution. If I look at your box, is it, you know, a less expensive alternative to traditional storage? Is it the operational model that you make it up, some combination of it? Do you have any metrics you can share? Yeah, sure. So, I'm happy to share the list price. It's the most, when we look at the most, it's really an operational model. We're coming at a, we are going after the enterprise space with price for an enterprise user. We start at 130,000 lists and there's about 40 terabytes of raw storage that people can do to map. That includes the SDN switch. So, as you grow your environment, that cost obviously lowers as well. What you benefit the most from us is not only the high performance, 180,000 IOPS per 2U, you also benefit from the no operation at all, operation of the model. You plug in, it's plugging though. You don't have to buy a huge array to start with and then grow that with spindles, you buy a small chassis and you grow when you need more capacity and or performance. So it makes it very, very simple. And that's really where we need to move. You need to be able to consume storage as a utility not as an extensive. Here's my three year plan for storage and therefore I need to buy this much. Buy what you need for next year, for your budget year and then you can easily grow beyond that. So Doug, you know, it's tough to get a foothold in the marketplace. Is there a specific use case or you know, as you, you've had these first couple of customers, you know, that have been buying, you know, what, what is kind of the hook for them and, you know, how are you trying to get that, that first start in the market? So, you know, as the product manager, I'm extremely interested in use cases, particularly because of what we do for a living. What we've seen the most, the top two ones so far has been another surprise in VDI. So a lot of customers are interested in VDI. It's also probably from a product manager perspective, the least exciting to be honest. It works really well, but it's not necessarily one that's difficult to solve. The more interesting one we've seen lately is Oracle databases and the low latency access. So our box was very low latency performance. And we've seen customers coming in and saying, I want to move my low latency parts of the database environment, specifically Oracle in our case, to your box and see how that performs. So that can increase overall performance on a solution. And that's probably the top two I would pick for now. And as we, over the next few months, I'm sure we'll add a few more that are a bit more different from the sale address of the world. Yeah, Doug, to be honest, from Wikibon's standpoint, we'd agree with you, VDI definitely is growing, but is not necessarily the most interesting. There's a lot of solutions attacking that. There's huge opportunity on the Oracle side. And if you talk from a financial standpoint, we've done a lot of research that says that this is where companies can save huge amounts of money because it's such a mission critical piece and can really chop down. So can you talk to us a little bit about at this show, what are you guys doing? I know you guys have a booth. You've got a bunch of folks here. What are you looking forward to at this event? Sure, so far the show, by the way, has been fantastic. It's extremely busy down in the booth. I'm extremely happy to be here. We do a few things. One, we look for design partners open style. So as we build our products, especially as a startup, we want to make sure we build a product that's most interested to most customers. We were looking for a couple of key design partners that are able to get early access, early access to our drops and hardware to see how that can provide value in their environment. Second, we want to have interactions with the customers and obviously show off our product. So we have a demonstration downstairs. We have these awesome new chalk whiteboards in the background be able to show and have really in-depth conversations, which so far has been fantastic. So we have product management, technical marketing and engineering represented here at the show. So it's a really good opportunity for people to get into the depth of the product and understand our strategy going forward. All right, so OpenSack Summit, I expect we'll see you guys at the VMworld show later this year. What are you looking out as a product manager? Do you always have to make your bets as to, there's always a million things you guys could work on. So where do you set? What are the top priorities? What are the big things? Is it real-time, big data applications or what are the hot things on your plate? A few hot things. Clearly OpenStack is one of them out of us. We will not be here and developing for it. VMworld support and VMworld support will continue as VMware is a huge part of the market we're going after. The other part is it's going to be databases. It's going to be a database on a scale. Next year we'll introduce physical NFS for everybody and that will be even more important to have that scale behind the scenes. What you'll see later this year is even greater scale. You'll see even greater performance tomorrow with our next major release. And that's going to be extremely exciting because that will bring us into new areas of where the sand traditionally has lived. Now you can actually support that with an NFS workload, which is quite fantastic. It's really going to be more enterprise-worthy. When you look at a solution, highly resilient, highly redundant, you want to make sure it stays that way and I wouldn't see us venture too far outside of that. All right, your solution, does it tie it all to any of the public code environments to use it for backup or anything like that? I have nothing to speak to at the moment. Okay, all right. So Doug, I want to give you the last word on this segment. I know you've only had a little bit of time to look at the show, but you've been keeping an eye on OpenStack for a while now. Why is this point in time critical for the future of IT infrastructure and what's happening? Yeah, so I want to again correlate back to Linux. I lived through Linux for 10 years of it and it took about seven, eight years for it to become an enterprise product. I see OpenStack moving with almost about three times that speed. It's moving very rapidly. And I think it's nice, right? Time for companies to take that big step in and start, not only dibble dabble in it, but actually start deploying it in production. We see that from our customers and that there's a lot of interest, but they're afraid of taking the next step and I think they need to take a step back and look at what I really want to achieve in three years as a big IT organization. And OpenStack is a fantastic opportunity to move them to the next level where they need to be. All right, and I got one last question for you. Coho data, so Coho are the same and they swim upstream. Is there something in the name or kind of the founding story that speaks to that? We originally would call it a different name. I won't even try to pronounce it on the name because it's really, we want to stay with Coho data. We obviously, we're going into scale-on-ass. You can think that's crazy. There's so many big players in this space. What would we do it? We fundamentally believe we have something that's totally different from every existing player and we're happy to swim a bit upstream before we ready to lay our eggs down in that basket and we're doing really well so far and it's been a fantastic ride. All right, well, the salmon always have a tough journey but when they get there, they're excited and everything. So Doug, really appreciate you coming here, OpenStack Summit, sharing the story of Coho data with our audience and this is Stu Miniman. We'll be right back with our next guest after this break.