 from San Jose in the heart of Silicon Valley, extracting the signal from the noise. It's theCUBE, covering OCP, U.S. Summit 2016, brought to you by OCP. Now your host, Jeff Brick and Stu Miniman. Hey, welcome back everybody. Jeff Brick here with theCUBE. We are live in downtown San Jose at the San Jose Convention Center for Open Compute Project Summit 2016. The third year theCUBE has been here and this is really where the cloud is. This is all the infrastructure players. It's hardware, it's iron, there's nothing ambiguous about it. There's a lot of big metal here. And this is where it happens, this is the cloud. And so we're excited to be here to cover it for the third year and really excited to have Cole Crawford back on. He used to kind of run the Institute and then he left last year. We had you on, you started your new company, VapeRio. So first welcome back. Thank you very much. And give us the update on VapeRio. Oh, yeah, thank you. And thanks for having me back on. We just released Open DCRE 1.2 today. So you can think of VapeR as a, as an OCB contributor, which we absolutely are. We've got contributed for submission hardware that sits on top of an open rack. You heard the Open BMC announcement. We have an endpoint for that. We have an endpoint for Redfish. You know, the company last year, I don't know if we were able to say it yet, but we were invested in by Goldman Sachs, Austin Venture Partners, really fun, really fun, exciting times. So Cole, I'm wondering if you can reflect back for us, you know, somebody that's been there since the early days, you know, when, when OCP started, I mean, a lot of us were like, oh, cool, new way to build servers. And now you look at all the different activities that are going on in the breadth of the ecosystem. Is this where we thought it would be? You know, what surprised you? What happened as you thought? You know, welcome through to some of what you've seen. Yeah, I think, you know, is it where I thought it would be? I don't know, but yes, in a way, because Open Compute was always supposed to be that unplanned garden, right? It was always, let's plant something and, you know, something's going to grow. And we'll leave it for the community to determine what sticks, you know, and what sort of the beautiful flower that gets produced is, looks like, et cetera. And if that means that Erzholzley's on stage talking about Google joining, then, then absolutely yes, it's what I planned. That's great. I was just reading, you know, Mark Zuckerberg, you know, did a big post on Facebook and, you know, does highlight, you know, Google joining. I mean, that's the big news so far this week. Big news. When you talk about, you know, what percentages servers might, you know, fall under OCP in the future? If you've got, you know, Facebook's there, Google's there. You mentioned Goldman on the network side. You got Equinix, you know, signing up to do this. You know, some big names here putting their weight behind it. You know, it's really impressive. So, you know, any thoughts on that? You know, there's in anything. And you guys, this isn't my first rodeo in open source land or in nonprofit land, cloud open with the Linux foundation, foundation formation committee for the open stack, right? I've been an open source, I've been in nonprofits for a long time and you're always going to have naysayers. You're always going to have the people to say, and I remember when Rackspace and NASA came out and said, hey, we're doing this open stack thing. How many people said, oh, that's never going to take off? Well, you know, the run rate for that foundation is gigantic. Look at what Open Compute's been able to do this year. You know, if you look at the companies that have joined and where they're finding traction, I see just massive amount of momentum and a massive amount of potential. I think we've scratched the surface of what Open Compute can actually be. So, Coles, maybe speak a little bit more to that. When you talking, you know, starting your business, you know, how do you think about open source? How do you think about, you know, monetizing things? How do you differentiate? How do you worry that, you know, some big guys just going to come co-opt it and take over? You know, it's a good question. I think this, and I kind of maybe come at this at a slightly different angle than most people. I care a ton about open source licensing. And if you think back to two years ago when we announced the OCPHL permissive license and restrictive license, I architected those licenses and, you know, got Facebook's blessing and then got, you know, the board's blessing on here's the right governance model from a licensing perspective on how to do open source and hardware. Patent left type licensing, which, you know, to date hadn't really been done. I think monetizing, it's purely about innovating. For a company my size, you know, and we launched, if you saw the news, Applied Micro and Vapor announced a partnership today on an alternative to, you know, sort of lock-in rack scale, you know, rack is the unit of scale. It takes a lot of innovation to do that. It takes a lot of engineering time to do that. And monetizing that, something that can be done, you've got to be smart about how that happens. You need to partner with the right companies and you need to engage with the community. I think the one thing that OCP has proven and other nonprofits have proved it in other areas of the industry, but in hardware land, partnerships matter, right? The, no company here has it all figured out. The world of IT has gotten too big and you look at how open compute is sort of chipping away at the gratuitously differentiated lock-in mechanisms of the legacy, I'll say legacy OEMs and vendors. And you're starting to see an actual software interface, hardware interface, you know, a fabric, if you will, of open. And I think that that is paramount to the success both of open compute and you mentioned the Colos, you mentioned others getting involved. I think it's paramount to the success of the data center going forward. Yeah, Col, I mean, that's core to what your company's doing. You talk a little bit, I mean, think about, you know, power density and you know, kind of the cooling that we need to do. You know, lots of different rack designs that you do here. Where are we with it? You know, I've been to some of those data center component shows and there's, you know, hundreds of companies doing cabling and different rack enclosures and it seems very fragmented. So what are we going to learn from the hyperscale guys? Are we going to see some massive consolidation and get flexible? You know, where is that part of the industry? Yeah, great question. You know, you look at the hyperscale companies and you see now a consolidation of SKUs. There's a smaller number of SKUs and I think the Intel acquisition of Altera is actually going to have something to do with this in the future. I think now with the parallel processing capabilities of the FPGA, I think that's going to be a very interesting, you know, future state for IT and data centers. But go back to 2010 when Facebook started Open Compute and they started it as project freedom and this was freedom from the gratuitous differentiation on how you manage the data center. So infrastructure management is a very key component of IT and how you look at your data center. And you know, I look at the data center space today and look, public cloud is dominating, right? Public cloud is absolutely dominating. We saw Google now and you know, I leave it to the industry to figure out, you know, when does an Amazon join, right? When does a LinkedIn join? When are the other big hyperscale companies going to come play ball? We don't need to reinvent the wheel here. This is not about not invented here. That's the wrong way to look at this. But we need common abstraction layers. You know, one of the benefits of OpenStack is it gives you this sort of common abstraction layer. We have to see that in hardware land as well. And it's not just about the physical interface. It's about how do I actually talk to my data center in context of my application? Because PUE, you saw Ashrae is dropping the PUE requirement. How do we get to performance per watt per dollar? How do we get to performance per watt per currency? Which is a metric that we should be using, right? We care about the workload. The Jeff Bezos is going to make sure that hardware is a loss leader. Let's manage our infrastructure in a common way through standard interfaces that are agnostic to Redfish or OpenSwitch or OpenBMC or any of the other sort of open environments out there on how you look at infrastructure. And let's get to a place of sanity so we can start focusing on the application. And as much as I love the innovation in hardware truly at the end of the day, it should be a loss leader. In many cases, hardware should act as a loss leader. The service is what's important. And monetizing that service, I think, is what's the opportunistic thing at stake here. But this is kind of interesting ying and yang, right? Back and forth, where before people had proprietary microprocessors and proprietary everything and then everyone kind of jumped to industry standard, which basically meant Intel. Then we had James Hamilton on from AWS a couple of years back at re-invent, who said, now we're actually going back the other way. We've got a big enough scale. We're building purpose-built hardware. But now we're back here at OCP and Intel's lead sponsor with Facebook and obviously data center businesses that are growing part of their business, Cisco UCS, et cetera. It's kind of going back and forth, right? And it seems, so are we swinging back to everybody here gets it and we can kind of build around this open standard? Hashtag pendulum of IT. So we, and we see this across the board. Look at how this world started. Big massive mainframes, right? Then client server, then superdoms and 15Ks and 25Ks, we call it big iron. And now it's about cloud and then there's private cloud and public cloud. Finally there's some equilibrium. There's a state of equilibrium that we found here. I think that's hybrid cloud. And that's, I've got some on-premise. I've got some remote. I might have some managed service. I might have whatever those are. But how do I get the telemetry? How do I get the situational awareness of my data center, regardless of where the workload is? How do I manage that? I finally think that 2016, maybe 2017 will be the year of dynamic and proactive data center environments as opposed to archaic and at best near time data center environments. Well, it also feels like everyone realizes that all these large customers have lots of different vendors. They have lots of different departments. They have lots of different applications. They have lots of different solutions. There isn't just going to be one vendor that owns the whole thing. And it was interesting, I think, Mark and the keynote for Microsoft saying you might be surprised that we're all about open source because we have Azure and if we're going to operate Azure, not everyone is going to want to put Microsoft-only stuff in there. So it's an interesting dynamic, kind of this acceptance of, we'll get our piece of the pie but to get more of our piece of the pie we actually have to inter-operate with others that are playing in the same sandbox, if you will. Right, co-op petition. And I think that that's the, if there's a phrase that I think people should be thinking about this world as, this is a bunch of vendors that compete in some areas but are complementary in others. And again, the NRE that goes into a lot of the engineering for this work, the capital cost of doing that inside of one company's four walls is astronomical at this point. So why not rely on an industry that wants to help you sort of rise the tide and raise all boats? Now, kind of the dark side potentially of open source and the enterprise side is management. So you've got a great perspective you've been at it for a long time. If I'm doing, you know, commercial kind of commercial version of open source but then the project start blossoming like, you know, many, many mushrooms in a dark, cold room and it happened in the bay. You know, how enterprises kind of keep that wrangled in while at the same time, you know, really trying to embrace because the innovation that comes out of open source is unparalleled. Right. Yeah, and management, I think, is one of the places where, you know, the world is going to quickly make up its mind about how it deals with orchestration. You look at the innovation on the software side. You've got, you know, for years, for almost a decade, you've had a strong story to tell around DevOps. You've got all sorts of platforms to run and automate the software side of your data center. You've got now containers and container management companies. You've got Mesosphere. You've got Docker and Core and Kubernetes, you know, CoroS with Rockets. And it's fantastic if you've got the capability to manage your application that way. These are cloud level application environments that you want to orchestrate. Where's the DevOps for hardware? If you're going to run a software defined data center, you actually need to talk to your data center. And so by having been a software guy and having been an architect guy, a platform guy and an OS administration person, and, you know, with open compute, sort of forced my hand at trying hardware. I think I've got a pretty interesting, you know, I certainly think I've got a pretty interesting perspective on hardware and just the data center as a whole. And I think that's key. You need to be looking at the data center as a whole because Colos, you know, any end user, they care about Cosper Cloud, but they also care about Cosper Square Foot and Cosper Kilowatt Hour. And they care about the efficiencies and the performance they're able to achieve out of the application. And I promise you, if you go to any Fortune 50 CFO today, you say, are you more competitive in cloud or on premise? You would say, I don't know 49 times. So, Col, we don't have a lot of time left, but you know, you've got an interesting story of how you came up with the idea for your current company. I wonder if you could share this, share that with our audience. Yeah, I mean, you know, aside from the software bit and the fact that I've been in, and my Twitter handle is calling the cloud, you know, for a couple of reasons, one of those reasons is, and you know, you skydive like I do and you want to train, I actually do this sort of semi-competitively. I started thinking about the efficiencies and the densities and just the way air moves in a data center. And air doesn't like to be pushed east to west. Air likes to go north, right? Hot air likes to rise. So, and I remember something that Mark Zuckerberg said at the Open Compute Summit a couple of years ago, you know, around higher inlet temperatures, and this all made a lot of sense to me, but what didn't make sense to me was the hot aisle. I've never been a fan of hot aisle, cold aisle containment. It just, why are we trying to move air this way? Yeah, the server's built that way, but guess what? It's because the rack was built for railroad switching and telcos in the early 1900s and we knew how to mass produce it, so we put servers in that fit that. We tried to think of this slightly different. How do you maximize the efficiencies and minimize the white floor space? Because, listen, co-location companies, data center companies, there's finite amount of white space that you get to put your data center infrastructure in. So how do you maximize that? How do you make it pain-free to do in-place upgrades, et cetera? So I got the idea of the vapor chamber, which is an open rack, open compatible OCP, open rack compatible solution, and we use the Bernoulli principle or the Venturi effect, which high pressure, low pressure zones, we pull air in and we pull it up in very small environments. It's actually got that idea one day, flying around in the tunnel of a skydive simulation, you train in these tunnels and it's the same vertical air movement. Obviously not the same as what we do, but that was sort of at least one of the inspiration. I mean, love the intersection of your two passions and pulling together those two cloud discussions. Passions drive innovation. Love it, that's it, passion drives innovation. So if you're not passionate about what you do, do something different, right? You'll probably be more successful because you can't compete with somebody that's got the passion if you don't, if you're just punching it for work every day. So, Cole, thanks for stopping by. Thank you guys. I knew you guys, you kept running up. Cole's always on the move, whether he's falling out of airplanes, perfectly good airplanes and doing tricks or running up to the cube. Thanks for stopping by. Thank you guys. We'll keep an eye on April 8. I'm Jeff Rick, Mr. Miniman. We're in downtown San Jose at the Open Compute Project Summit 2016. We'll be back with our next guest after this short break. Thanks for watching.