 in the heart of Silicon Valley. Extracting the signal from the noise. It's theCUBE, covering OCP US Summit 2016. Brought to you by OCP. Now your host, Jeff Brick and Stu Minow. Hey, welcome back everybody. Jeff Brick here at theCUBE. We are live in San Jose, California, the San Jose Convention Center at Open Compute Project Summit 2016. It's our third year bringing theCUBE here. This is where the cloud lives. This is the infrastructure. This is the hardware. This is what's powering the data centers. And it's a lot of excitement. If you like geeky stuff and you want to touch it on like software, this is the place to be. People walking around with cards and there's boxes and servers and all kinds of fun stuff. We're really excited to be joined by our next guest, JR Rivers from Keenlist Networks. We saw you here actually two years ago. I feel like we've seen you a couple of times since then but you were here two years ago and it was still number five. Even though it's about 2014. So give us your perspective. You've been in the Valley a long time. You've been with a lot of great players. What's your kind of take on what's going on here this week? Well, when you look at Open Compute and how it's evolved, right? It started off in kind of the warehouse areas of Facebook, a small invite-only event, not that many people. I think they were trying to figure out what it all meant, a little bit of recruiting, a little bit of enablement, a little bit of supply chain management. And it's evolved itself to a point where customers are really trying to get their heads wrapped around. What does it mean to have a supply chain that looks kind of like a big player, right? They know that a Google or a Facebook has kind of this ODM direct relationship with their suppliers and they want to know what does that look like? And they've been looking over and over again, like you said, years and years they've been coming to these Open Compute summits saying what is this about? So on the floor you see suppliers that are people like Melanox, Chips and Systems, Finisar, something like selling optical cables and modules, and they're saying hey, this is what everybody's building mega-scaled data centers out of. You can buy this from us directly. And so it's pretty cool to watch customers really latching onto it, recognizing that hey, this is in the realm of possibility. I've heard the same thing four or five years in a row. And what's been interesting to us is over this past year, we've had numerous customers that have said we're building a new data center and it's all open compute technology. Pretty cool. Wow, that's amazing. So JR, we used to see that there was a huge chasm between what the hyperscale guys were doing in the enterprise. I mean, I think you're one of those guys that helped build a massive infrastructure and the typical enterprise doesn't have JR and a team of PhDs to help architect this. So does OCP help bridge that gap? Help us kind of square that circle of, the infrastructure has to change but the applications aren't the same. The operational model is very different. So is this still for like the big 20 guys or how far down market does it come and how does that operations change? Yeah, I mean, we're lucky to see it on the customer side and it goes all the way down market. I mean, we have little tiny operations like a library here or a school district there. They're able to leverage OCP technologies, OCP switches, OCP servers. So it doesn't really stop at any size or scale. I think the big difference is the enterprises had been taught by their suppliers for a really long time that all this stuff was really hard. There's a lot of black magic in there. Don't worry, trust me, I know how it works. And by coming to a show like OCP and Compute Summit, they're recognizing, hey, there's really not that much black magic in here and it's not that hard to put together and I'm being told by numerous people and show an example after example, that it's not that hard and they're just becoming more comfortable. Yeah, so network has been a big focus of the news and the keynotes this morning. This morning in the keynote, they kind of walked through the stack and talked about the disaggregation of hardware and software, which Cumulus is heavily involved in and then kind of moving up the stack. Well, where do you see things? Where's the ecosystem, the partnerships? Where do you play? Where do you partner? Where don't you partner? Right, yeah, so for us at Cumulus, we're all about software, right? Whether it's kind of a box level OS that helps you run routing protocols and make the box do things or management layer software that wraps around those boxes and allows them to do interesting things in customer's networks. So that's kind of our world. We partner really well with people that make great hardware, whether it's a company like Mellanox with a Spectra basic and some of their optical modules and cables as they're moving forward, someone like a Dell or an Acton Edgecore or a Quanta systems. People that allow customer's access to high capacity hardware, whether it's compute or storage or networking, those are our natural partners. On the software side, we partner with a lot of ecosystem people. We have a great relationship. Actually, I don't want to name too many people, but anyways, we have great relationships with a lot of software partners as well because we're less about what code we write and more about what we can pull together solutions to allow our customers to benefit. So JR, a couple of big announcements from some big players here. So Microsoft's been involved in OCP for a little while, but Google is participating. What's your take on that? What's real? What's the motivation for them to being here? Do you think Amazon's going to show up next year? Right. Slightly surprised that Google showed up to be blind. They develop a ton of technology. They consume so much of it they don't really need to expand the ecosystem around that. They get great supply chain by themselves. Maybe a little bit of it is get them giving back. Not a traditional Google-y thing to do. A lot of times they keep technology to themselves. So I wish I could tell you more on what their motivations are, but I would love to drink beers with somebody to figure that out. The Microsoft cloud, sweetly they call it Sonic or whatever. Sonic, yeah. Generally pretty cool and great indicator that yet another web presence says to someone like Cisco, hey, your software's not that interesting to us. In a super really weird way, it's a great validation of what we're doing at Cumulus because it looks a lot like Cumulus Linux. They copied a ton of stuff from us. We've met with them. If you break it down, it's a lot of our components. Their GitHub repository contains a bunch of source code that's got our copyright on it. So it's a huge validation of what we've done that they've seen that so useful that they're using it in their production data sets. They're not a customer, as you can tell. But they're leveraging the technology. I'm wondering if it has something with Google in kind of support of the cloud project, right? To be more in kind of the world of the enterprise suppliers and maybe it has something to do with that. There are events coming up. We covered their cloud launch a couple of years ago. I don't know, just speculating. I think it's a fair speculation, but I would argue that it's probably more PR than technology if you get what I'm saying, right? I mean, if you look at something coming like Google and what it's going to mean for them to be an enterprise in a cloud, it's going to be less about, hey, I can give you a VM at the lowest cost or I can give you a container for almost free and more about I've got this really great enterprise class application that will solve a meaningful business problem for you. Right, yeah, the other piece, Brian Graceley, our cloud guy at Wikibon always said, the guys that you see at the keynotes at every single show, that means they're recruiting. Right, yeah. So it's all, you know, there's a limited pool of talent, they're all going after these various pieces. Let me ask you, JR, your customers, you know, how much is open source something that they're like, it must be open source, or do they want to participate? What is that dynamic of open source and what you're seeing from the customer base? Yeah, that's, obviously there's a spectrum like with anything else like that. What we're seeing increasingly is, it's kind of a little bit less about open source and more about the ability to be flexible and move around, supply chain, control, ability to influence the outcome. So some people are like super religious, I want to have open source because it affords me all of those things we just talked about. Some of it's more of, I just want to know that what's under the covers can be replaced, worked on, enhanced, outside of whoever I'm working with, I don't like that world where someone like, you know, a Cisco's wrapped me all the way up and I don't have any degrees of freedom to move. Yeah, it was interesting, I was watching one of the keynote presentations, they were talking about, you know, just the super high band width, you know, that's coming in the future, and they said that, you know, we might even be getting away from connectors. You know, it's just going to be built onto the board. I know something you and I've talked about many times in the past, you know, just cabling and optics is kind of this dark art that like most people don't talk about, but it's a multi-billion dollar industry. Do you see that changing? The physical mechanics, I see it changing, but I mean, just let me flip it around on you, like when you look at the 100 gig technology that's going to start, it's coming out this year, I mean, you're looking at 100 gig switches that are going to cost customers like 10 grand, 32 by 100 for 10 grand, just kind of use your brain and roll back. How long ago was that the band width capacity of like a half a million dollar chassis, right? About five years ago. How long ago was that the total band width capacity of a telco pop seven, eight years ago? How long ago was that the total band width capacity of a telco 15 years ago? That's fucking amazing. Yeah, that's amazing. I mean, the general rule of thumb we used to have is that if I could get 10x the band width at 4x the price, you know, it was kind of that no-brainer that I would definitely, that was when you saw the inflective, because it always took so long for us to get the adoption of, you know, I mean, 10 gig was ratified back in 2002. We're, you know, still, you know, not a 50% adoption on the server rate. The core, the backbones are all doing well. So, yeah, I mean, you know, there's discussion 100 gig in beyond now, but boy, on the server side, everybody's talking about 25 gig. It's not that sexy. But, you know, these transitions are tough. You know, we run into some of the limitations of the economics and what's happening there. But what's happening is with, you know, kind of forums and foundations like OCP, some of the stuff we're doing at Cumulus, and other people that are kind of in that mix and ecosystem, is they're starting to remove cost as a barrier to capacity. Kind of across the board, right? You can get a phenomenally operational network, great hardware technology, and still, like, continue to run your business, right? And hire people to actually operate the year. Yeah, so one of the discussions we've had at this show is that, you know, networking becoming part of the discussion helps to kind of get it out of its silo. And I think, you know, one of the cool things about Cumulus is, I mean, you guys are Linux. You know, it's the way I operate it. You know, if I understand Linux, I can, you know, manage all your devices. Do you see networking kind of coming out of the bowels and be just part of a general, you know, infrastructure? You know, where do you see that dynamic of, you know, networking in general IT? So, that's a great question. You're 100% aligned with, A, what we see happening, and why we've taken the path we have. You know, I've been in networking for a really long time. We could have written our own proprietary networking stuff that's really cool. You know, like maybe it's an HP open switch. Maybe it's a, you know, a risk to EOS. But we recognize that Linux is going to be the foundation of data center going forward, storage, compute, networking, all of those pieces. And being able to leverage them in concert is super important. We have a bunch of customers that use some overlay technology that we built. It's kind of a super, we call it lightweight network virtualization. Think of it as kind of like, you know, baby step, nice era. And they use it on the switches and the servers because they're all Linux, uses standard Linux technology. They're putting on storage boxes now. So, way of them building a reasonably cool network that's, you know, way ahead of what is generally available all using Linux building blocks. Really, really cool. Kind of interestingly enough, our VP of engineering either has done a talk or is doing a talk. It's called Linux Networking is Awesome here at OCP. He's doing something really, really cool. He's got a Broadcom based switch and a Melanox based switch in a network together. That's not the cool part. They're running MLAG. So, if you look at like classical vendors, you know, basically they all rely on kind of internal header stuff. And MLAG is only compatible between like this product line of this vendor and not compatible across the multivendor and all that. Because we've based our MLAG off all of the Linux constructs, it works, Melanox and Broadcom together. Not same vendor, not same silicon, not same anything, just Linux. No, that's pretty cool because, you know, we always talk about interoperability when it comes down to it. Yeah, you need to do something different because customers always ask for it, but it's usually not that simple. So, we have to, you know, to make something simple is really hard. Right, exactly. That's why we're really proud. So, Jer, I want to shift gears a little bit because we always hear networking as the lag between compute and store and kind of the rapid growth of the cloud. In your experience with customers, is the network catching up to the demand or are they getting this new capacity that you said they just went through the steps of this massive capacity increase? Is that getting ahead of their demand and are they starting to come up with new applications, new solutions to take advantage of that? What's kind of leading which? Yeah, that's an interesting question. I think you're starting to see this age where like at 10 gig to the server, you were kind of barely starting to catch up to what the servers were able to support a couple of years ago, right? So, you're like one gig to the server or maybe a couple of one gigs to the server, that was dead at almost 10 years ago. Like when I was at Google long ago, one gig to the server was squashed, right? It was way over provision. Sorry, over committed, over subscribed. So, getting to 10 gig kind of helps you out. And now all of a sudden with this advent of the 25 and 100 gig technologies, you're starting to open it up and people are envisioning operational environments that are way different than what they did before. We're seeing customers doing this crazy stuff where they're doing container deployments, they're running routing protocols on the host and the way they do mobility is the container has a static IP address just pops up, gets advertised in the network, storage shows up everywhere, there's no concept of placement of anything and you have these massive data centers that are running that way all because they have a phenomenal amount of interconnect capacity. That's awesome. So, JR, can't let you go without asking, what's kind of catching your eye these days? What kind of things as you look out, do you think people will be talking about throughout 2016 that maybe they aren't today? Right, it's going to sound slightly lame but I think it's going to end up being true and that IP storage and what are the right solutions, the scalable, tangible, high-performance IP storage solutions, that's going to end up being the big discussion. People are recognizing that it's possible and it's viable and they're starting to run up against some of the bottlenecks of it, you know like, Seth, let's pick on that, right? Great technology, got some limitations, we have a lot of customers using it, bunch of them are like tweaking it, like they're writing their own patches to make, to come up with issues or to overcome issues rather, you're going to start seeing people continue to beat up against that until IP storage becomes the foundation of what people do and I think that's going to happen over the course of the next year. You're laughing, do you agree or disagree? I mean JR, you and I have history working on these technologies, you know, I feel like we've been beat our heads on that wall for a while, my take, I mean there's certain applications that just, you know, they're using Ethernet, I mean everything's using Ethernet, when I do SaaS, I talk to most service providers, it's going that way, you know, hyper-converged infrastructure is not a sand environment, so yeah, I think that that movement is happening, but I think if you're talking about fiber channel as, you know, versus IP, I feel like that discussion happened two years ago. Yeah, especially with fiber channel versus IP, it's really just about- Helping storage, you know, networking solves some of the storage problems, is what I'm saying. Or how it systemically solves the storage problems, right, when you look at a technology like Seth, you know, it's like in the old days, like when we used to do this stuff, right, someone would assume I'm going to have an IP storage target, and I'm going to have a host that's going to talk with that target, but with something like Seth, you have the local caching layer, that's super sophisticated, right, that enables IP storage to be pragmatic. And I think that's what you're going to start seeing is, given an IP network, a super high capacity IP network, how does storage live in that world, where it's redundant, scalable, movable? Yeah, and some of your partners, you know, definitely we've been talking to about, you know, technologies that are doing this, it's definitely, yeah, good opportunity, networking storage to add tons of issues, and especially if they can work together to fix it. Right, exactly. That'll be goodness for the industry. I think it'll be fun. All right, well, we have to leave it at that, J.R., as always, good to see you, I'm sure we'll see you here again next year. Absolutely, sounds good, thanks a lot guys, appreciate it. J.R. Rivers from Cumulus Networks, I'm Jeff Frick with Stu Miniman, we are live in San Jose, California, at the Open Compute Project Summit 2016. Again, big shout out to Pica8, and Micron and Melanox for sponsoring us to be here, without their support and the support of our other sponsors. We couldn't take theCUBE to all these events over 80 last year, and bring you the insight that you're looking for. So again, thanks to them, we'll be back after this short break with our next guest. Thanks for watching.