 From San Jose, 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 Frick and Stu Miniman. Hey, welcome back everybody. Jeff Frick here with Stu Miniman. We are at OCP Summit, Open Compute Project Summit. It's really where open source meets hardware that drives the cloud. First started by Facebook a number of years back to really open up the Kimono, if you will, to their hardware infrastructure. Because as Stu said, it's not their core business. They want a lot of people to participate and really leverage their learning about hyperscale, bringing into a broader community that helps everybody really drive the cloud innovation. That's so important to everything that's happening in the technology world. And we're really excited to get a return visit from my guest from last year, James Lau, the founder and CEO of Pika8. James, welcome back. Glad to be here, gentlemen. Absolutely, and thank you again for your sponsorship to help theCUBE come to events like this. It really means a lot to us. So what's happened in a year since last we saw you? Well, I think the open network and the Y-Box network has made tremendous progress. You guys can see it over here. The whole OCP now is centralized on the networking side. Everybody is trying to figure out what's going on with networking, how to solve the networking problem. The biggest change we have seen in the last year is the ecosystem is coming together. And this goes really well with the announcement this morning about Microsoft is joining the OCP with the networking contribution as well. So we're seeing the Y-Box switching and the SDN are coming together with the ecosystem. And that really can solve the problem. And the ecosystem includes the application, includes the controller, includes networking opportunity system, includes a lot of platforms. So we're seeing over here that a lot of players are here to really push this forward and really solve the problem. Well, it's a great opportunity for you because people often talk, you know, there's Compute Store and the networking that the networking was really the laggard of those three in this kind of big move to big data in the cloud. So a terrific opportunity. Exactly, so I think the networking has been an independent industry for 20 years. People are trying to solve the Y-Box server problem, Y-Box storage problem, and they are inventing all kinds of software on top of server and storage. But networking always was an independent silo until three years ago with the SDN with Y-Box and people start to open up the networking and say, we need to solve this problem. Yeah, so James, I thought it was interesting to keynote. First of all, you were right. Networking seems to be the main topic talked about this week, but they walk through kind of the networking stack because it's not just one piece. It's not about, oh, let's change the hardware and you know, wedge is kind of interesting. You know, Microsoft has, they call their switch abstraction interface and then they have Sonic sitting on top of it. They've got a whole bunch of partnerships and a lot of environments. Can you talk about, you know, what's the challenge that Pika8 is looking to solve? Where do you fit in that stack? What are the partnerships that you're having, you know, as it relates to OCP and beyond? Absolutely, I think in the past three years, people have solved a lot of problems in the networking. So first of all, people start to be able to separate the networking hardware from the software. So this was a big achievement, but because once you can separate the hardware from the software, people can start to do the innovation. And then we start to realize you can separate. So I'm sorry, just a question because the white box tends to be some people don't understand it, what it is, where it is. Sure. I was talking to some Wall Street people and they said, you know, adoption of quote, white box switching, it's been flat for like two years. So can you explain for our audience a little bit about what it means to be, you know, white box, you know, where are we really seeing, how do the, you know, solutions that like, you know, Dell and Hewlett Packard Enterprise, are those white boxes, are they counted as something else? You know, what was with that piece? I think white box itself has a lot of different definitions. Definitely it's confusing, but definitely we see bright box as part of the white box as well. And when they count the white box, they might not be thinking about that. But if you think about the adoption of a white box, there are three segments, really, really interesting in white box. The first one is data center. Everybody knows that service providers needs white box, they can reduce the cost, they can scale, they can optimize the traffic. And this is the segment that really knows about how to use the white box, not only on the server and storage, but also on the networking side. But if you move a little to a different segment, enterprise side, enterprise has been moving a little bit slower than data center, because they tend to be more conservative about what kind of service they want to use. So if you ask the enterprise servers or enterprise customers, they will think about, hey, white box, maybe it's not ready yet. But we're seeing very encouraging sign this past year, we partnered with the HP Enterprise on going to enterprise market. We saw a lot of interest over there. People look beyond Cisco, people look beyond Dell, and not Dell, I mean Juniper, those kind of providers, and looking into HP and Dell as alternative providers, not because it's cheaper, but because they provide certain flexibility, they can optimize their environment, reduce the operation cost, and then provide the service. There is another segment that people didn't realize, it was really interesting for networking, which is in the OEM business. In the past, networking has been an independent silo, and everybody building storage or security, they have to build around networking and compromise with the networking limitation. This is the first time that they can just take a white box, optimize the software for their application. So we are talking to a lot of different storage vendors, for example, they really want to optimize networking for their storage, instead of changing storage for networking. So we just announced the... Wait, wait, wait, say that once again, that's a pretty powerful statement. So in the past, they would change the storage based on the network, now you're saying they're changing their network based on the storage. Exactly. And this is the power of a white box, because in the past, for example, if I want to build an ice-cold cloud, I always have to think about, hey, networking cannot do this, I'm going to drop the packet, I'm going to lose the route, how do I recover from this kind of things? But now they're telling us that if I can build the white box into my OEM, if I can OEM the white box and put it into my storage cloud, how do you change the networking so that you can guarantee me you are not dropping the packet, you can deliver the packet to the rice server? And in this way, not only they have lower cost, they actually work around all the limitation they have on the networking side and build it as their differentiation. That's one of the market that people really underestimate how much networking can do to improve the performance of the application. Okay, so James, you know, we talk a lot about the hardware here at OCP, but if we change the hardware, really it's the software that really needs to understand, manage, handle that, and of course, you know, PK-8 is a software company, can you talk a little bit about, you know, where we are with SDN and kind of the value prop that PK-8 puts there? Right, so in the past three years, we've been focusing on several goals and we achieved them one by one. The first one is to separate the hardware from the software. It was a difficult stage step, but with OCP, together with OCP, we start to achieve that. People start to realize, hey, hardware, I can really pick any kind of hardware as long as I control the software. And then the next step we took was to separate the control plane from the data plane, which means that if I really want to change the traffic, I'm not limited by the control plane. But the next step is actually the hardest step and we can see a lot of different approaches trying to solve that problem. The last step or the last black box was actually the ASIC itself. There will be different ASICs, each one of them fit for different applications. And people start to think, hey, maybe every ASIC is different. It's the same. The truth is that every ASIC is different. So if you think about Brocon, for example, Brocon is very good at scalability of the performance. They fit very well into the data center. But KVM chip is very good at flexible pipeline, which means that if you want to optimize for your storage or security, they will be really good fit. And Dune chip has a big buffer if you want to optimize for your storage, you really want to have a big buffer in your networking. So every chip is a little bit different. But now in the past, nobody saw the chip. So now how do you open up the chip so that people not only can program the chip, also see the differentiation of the chip, bring it to the application level. That's the difficulty. Wow, so one of the real challenges, you know, most users and especially even the hyperscale guys have this on the one hand, standardization and simplification is what helps drive my economies of scale and allows me to manage it so much simply. But I'd like to be able to have flexibility. And we look at somebody like Facebook and some of the guys they add up on stage, I mean, they're PhDs. Facebook has this team of PhDs, Google has teams of PhDs. Do I need a team of PhDs to be able to leverage some of these new ASICs? Is this something that we're going to see in the broad markets or can you help unpack that? Sure, so I think there are many ways of solving the problem. You can definitely host a bunch of PhD and try to solve the single problem you have. But in the meantime, if you are a small house where you are enterprise, you really don't want to host a PhD and turn yourself into a networking vendor. You definitely want to look for a solution that really you can get a turnkey solution and really get the solution that solve your problem. So that's why building the partnership is really important. PICAI has the capability that really bring the differentiation of the chip to the application level. But we're partnering with many different partners. At the application level, we partner with, for example, HP. They have an app store that with a lot of turnkey solution. People can look into that and say, I want to turn this one. On the platform level, we have a lot of OEM or ODM partners like HP or Contran, we just announced it in Europe, or Edgecore that's trying to solve the data center problems. So these vendors provide not only low cost but really high performance platform, high reliability platform to customers. But then you get to the next step, look into the ASIC and say, what kind of ASIC can I use to solve my specific type of problem? And you really don't want to just hire a bunch of a PhD and figuring out how to turn those on. You can just rely on PICAI and then go with these partners, we'll present a turnkey solution and solve the problems. So James, I'm wondering if you can comment. There was big growth of the ODMs. And OCP kind of enables the ODMs to be able to reach broader markets. Yet the OEMs are also stepping forward and getting part of OCP. Do you see the line between ODMs and OEMs blurring? Where do you see that dynamic going? I think they are definitely blurring, but there are some fine lines between the two. I think it will eventually differentiate each one of them to go to a different market. So on the ODM side, they are much more flexible in changing the hardware customized to the things. And for big data center like Facebook or LinkedIn, they really want to have customized network gear that put it into their specific type of rack. So this is the market that ODM will do very well. But in the enterprise side, they need reliability, they really need a good support on the hardware side. So people like HP will go really well into the enterprise side. But don't forget, even though everything is looking the same, there are always rooms for innovation. So for example, HP can easily create a turnkey solution integrated with a specific kind of ASIC they want to use and presented to specific type of customers. So for example, the storage network, they need a deep buffer. Who can really create a deep buffer gears for enterprise? You not only need the gear, you need the chip, and you need the software to drive it too. So HP is in the good position to solve that problem. But if you go to data center, they have their own way of solving that problem. They don't really need a vendor to come in to solve that software problem for them. They will just go with the ODM to solve, to get the low cost, the high performance scalable platform to solve it. So we have a great interview, James, with Martin Casado from Nassira who with VMware now is at Andreessen Horowitz. And it's a funny story, he's talking about SDN, the journey to SDN. He says, every year I come on, we're getting closer, stay with me, stay with me. It sounds like we're really there, but I wonder if you can share your perspective on, how do you leverage open source and still maintain innovation and competitive advantage? Those things to the outside world maybe look like they're opposed. How do you leverage the one and still find that opportunity to be innovative and separate your products from your competition? Exactly. So I think in a way that if you look at the market, there are only very few users can really download everything from the open source community and then put together something and then feel comfortable putting into production environment without knowing who is going to support it. Google can do it, Facebook can do it, but if you go to an enterprise and say, hey, you want to do it? They say, no, this is not something we can do, right? They need somebody to put together the solution, but open source indeed helping everything move forward. Why is that? Because first of all, if you have an open source, the hardware people know what kind of software they can run on the hardware. So they tend to unify the hardware platform and that will help us too. And then secondly, for us we have our key differentiation on optimizing the networking traffic for a vacation. This is something that open source cannot do for customers, but in the meantime we can leverage open source to solve some of the problems that everybody has already solved. So for us, layer two, layer three, for example, is not a key differentiation. People need layer two, layer three to scale the traffic? Yes, but they need the support. They don't need a new layer two, layer three protocol to tell them what to do. But in the meantime, you can really create some of the differentiation by telling people, hey, you can optimize this for your application. I can optimize the buffer usage, for example, so that you don't drop the packet or I can do the traffic engineering so that you don't follow the layer two, layer three rules. All right, James, so they're telling us we're out of time, but I'm going to give you the last word. We saw you last year. We're going to see you next year. Give me real quick three things you're working on for the next 12 months. Boom, boom, boom. I think the SDN and Warbox are here. They are ready. We're going to see growth, good growth of the revenue and customer base. We will continue to build a partnership. You are going to see more and more vendors partnering with each other and we definitely are reaching out to all the players. And we're going to see a lot of use cases from first mover customers because this is going to be the exciting time. People understand networking is the barrier and we have to work together to solve that problem. All right, James Lout, founder of PGA. Thanks again for stopping by. Thank you. Jeff Briggs, Stu Miniman. We are live in San Jose at Opel KPU Project Summit 2016. We'll be back with our next guest after this short break. Thanks for watching.