 Live from the San Jose McCannery Convention Center, it's theCUBE at Open Compute Project US Summit 2015. Okay, welcome back everyone. We are here live in Silicon Valley for the Open Compute Project Summit 2015 or OCP Summit 2015. That's the hashtag, Join the Conversation. This is theCUBE, our flagship program where we go out to the events to extract the synth and the noise. SiliconANGLE Media Production, which includes SiliconANGLE, Wikibon, and theCUBE all under one umbrella, bringing all the signal to you. You can hear all live in Silicon Valley where all the action is. This is where the innovation is. I'm with Jeff Frick, my co-host, our next guest is James Liao, founder and CEO of Pika8. Welcome to theCUBE. Thank you, nice to meet you guys. We were talking on our intro. This is where the action is right now. All the smart monies here. You can see all the smart VCs that I know that what's best in the business are out here. There's real technical innovation going on, but a shift and an inflection point happening at the same time. Software defined infrastructure, right? Everything software defined, and you have embedded concepts, systems on a chip, intelligent software, adaptive like fabric. These are decade old concepts now brought forward open source. So this is greatness. Now everyone always says, oh, that's great, but the network's still the bottleneck. So what the hell's going on with the network? Is it going to get faster? Virtualization, SDNs on the horizon? Give us the update. Where's the innovation of the network? So basically I think that networking is going through a big paradigm shift. In the past 20 years, we have seen a lot of changes in the computing, in the storage part. The hardware is commodized, standardized, and then people bring the software, and then eventually software bring the virtualization. So there are a lot of changes, and now rolling clock forward 20 years later, now with the technology we have in the data center, you can create a virtual machine within 30 seconds. You can attach one terabyte of storage within 30 seconds, but once you are done with the virtual machine and the virtual storage, how do you deploy them? If you want to deploy the virtual machine into your data center, you have to go through networking. And all the networking right now is manually managed, which means you have to create your firewall policy, your security policy, your low banter policy, all your routes, how does traffic go into this machine, how does traffic go out of the machine? So the change itself, by deploying the virtual machine which takes 60 seconds to create, now it takes one day to configure. That's how things slow down. So in order to solve that problem, we just have to think about what we have done with computing and storage. The technology we bring to storage and computing is standardizing the hardware, bringing more software, creating more application, creating more innovation to the software so that we can create a virtualization of the networking. So I noticed you guys did a crowd chat with Scott Rainovich, one of the friends of theCUBE, also a guest co-host, Scott, are you watching? I hope the slopes are okay. It's 65 degrees, you said, up in Bozeman, Montana. Good friend, but he had a very interesting conversation. One of the top voted questions was, who's involved in this white box food chain and what does pika aid do? And hashtag SDN, NFV, NFV implying service provider. A lot of stakes on the table right now in terms of high stakes poker. Exactly. What's going on? What's the white box mean and what's the big prize? What are people fighting for? So I think the networking is good. The paradigm shifting in the networking industry brings a lot of opportunity to all vendors. There are a lot of segregation in the supply chain. So people start to realize, I can separate the hardware from the software. People start to realize I can separate control plane from the data plane. I can separate the physical network from the virtual network. So by doing different segregation, you start to bring in more vendors into the game. And on the white box side, for example, if we talk about white box, now there are a lot of vendors from Asia, from Europe, from US that can create a white box, which is very efficient, very high performance, very low cost, easy to manage, easy to program. So vendors like HP, Dell has already joined the game. Even Juniper is announcing that they want to be part of the game. And besides that, from Asia Pacific, you can see a lot of vendors like Quanta, like Wistron, like Celestica, and Alpha. This kind of networking vendors used to be doing a lot of manufacturing for big name-brain. They are now realizing, hey, this is the opportunity, we can come out to start doing the game. So on the supply chain side, on the hardware side, I think it's very open, and we see a lot of vendors join the party. On the software side, we're seeing a lot of changes on the software side too. Operating system side, you got Pica 8 in the game, you got Cumulus also creating a lot of innovation. But most importantly, on top of that, we see tons of vendors, dozens of vendors trying to create new innovation to change the efficiency of the network. So this aggregation is interesting here, right? So you're basically isolation allows for inefficiencies that take costs on manufacturing. And then the innovation can come in at any vector, software or other hardware component. Exactly. Okay, so let's step back a second. So for the average enterprise out there, who's got some money to spend, they have challenges. What's the core problem that you guys are solving? Is it the scale-out, scale-up combination? The scale-out's been a great thing, but they're used to scaling up, the old school vendors, which locked a little more general purpose. Can they get best of both worlds and how do you guys help the network solve this problem? Absolutely. I think the real thing people try to solve is that in the past 25 years, networking is basically just one solution. Cisco defines the architecture, everybody follows the architecture, you can replace a couple of top of rack, but you still follow the same architecture. And what we're hearing from our customers is that, hey, I'm not running the same application as other people. I have my unique need. Some of the people are running host service, so they got a lot of tendency inside the network. Some of the customers that run mega-data centers, they have to scale out, right? And some of the customers, they have only a few departments inside their organization, but they got a lot of security policy they have to apply. And some of the customers are telling us that networking is very important to their infrastructure, so they want to make sure it's a zero downtown. And in the past, you only have one recipe. Cisco tell you, I put this big core in the network and then you do active passive, right? And everybody do the same way. However, if you look at the application, everyone is using different applications. So it doesn't make sense to follow the same recipe. So do we want to do scale out or do we want to do scale up? That's actually not the very core of the question. The question now is that if I have my need of solving certain things, do I get the tool? Can I have the freedom to use the tool to solve the problems? This is very similar to the mainframe to PC change. In the very beginning of the change, mainframe has much more computing power than PC, but mainframe doesn't have any flexibility. You want to use IBM mainframe, you go to IBM, ask them to create an application to solve your problem. You want to... IBM System Z has Linux support now. We could argue that for the big banks, they can get best of both worlds. Exactly. In a single threaded environment. Exactly. Not necessarily multi-threaded. But 20 years ago, it was a big change. When people started to realize PC is much more flexible. I can write my own application on top. That's a big change. I think we are going through exactly the same change right now. The networking users start to realize that my need is different from everybody else. Let's break that down. I want to get your perspective, because you brought up a good point. The mainframe kind of triggered me a little bit, because we're covering the Z thing, which we like the IBM new Z stuff, but it's different. It's not your old grandfather's mainframe. Correct. And it's also single threaded, and their workloads very specific. You mentioned workloads. This isolation and this aggregation is interesting because now you can actually have flexibility in workloads. Talk about that dynamic, because whether it's NFV for service providers or something else, workload diversity is a big issue, right? Exactly. Isn't that fundamental issue right now? Workloads? It's a very important issue. So if you look at the workload, a lot of people, they run different applications, generating a lot of different workloads. For example, if I take banks as an example, in the daytime, you're doing a lot of trading. Your workloads are a lot of small packets running inside the network. Latency is the most important thing. But in the nighttime, you have to backup your database. Your trading database the whole day, you have to back it up. Now every package is big. And by the way, if you miss your window for opening trade hundreds of millions of dollars are at risk. Exactly, exactly. And if you have some application. Minutes or years in savings, right? Exactly. And if you have an application, for example backup application, they just screw it up and start sending traffic in the trading time. You probably lose millions in minutes. Hundreds of millions. Right, exactly. That's the consequence. So these are the things that you have to figure out. My networking has to support my application. So in terms of workload, when I look at the workload, most of people only see computing, they only see storage. They don't really know how networking is working inside this whole networking thing. So if you have a workload, specific workload, and you can analyze to the level. So here's the important thing. Not only you have to provision your workload on your networking, you have to know how these workload is performing inside your network. So you've got to have a cycle to say I can provision it and I can create, I can collect the data. So you need the monitoring and you don't want to compromise latency because that's the other one, Mitch Byer. Exactly. It's an interesting dilemma, right? Right, and you don't want to create more workload for humans, right? That's another thing. You don't want to say every time I program a network, I have to add a couple of network engineers to program a network. So what is the best way to handle this? Whatever we have done in the past 20 years to virtualize that computing and storage, whatever lesson we have done to reduce the CAPEX and OPEX, those are the things that we can now consider how to do it in the networking sector. So how do customers do it? Obviously they can't just turn off the old system and flick on the new and we do a lot of big data shows and we talk a lot about the journey, right? How do you get there? Where do you start? How do you find some early wins in some success? So within your customers, how are people slowly migrating over to a more software-defined network? That's a great question. We get that question a lot. When people talk about SDN, the first impression they get is that, well, SDN is going to completely change my network. So I got my current network running. I cannot afford to bring down my network and create a new one. So there is no possibility I can migrate to SDN. The fact is that there is an easy way to migrate to SDN. The first thing you want to consider is that through, I just mentioned that there are a lot of disaggregation between the different layers, right? Power versus software, control plan versus data plan, virtual network versus physical network. So most of our customers start to realize one thing. The first step they can easily take is to bring in YBox. You bring YBox, putting a different software on top. Now you got the freedom that you can run different software. However, bringing YBox itself has a lot of challenges as well. So you need some software company, like Pica A, to help you to migrate to YBox. So our first step to our customers is always trying to make YBox easy to use. And is that a carve out based on a workload? Is the carve out based on the data center? I mean, again, getting started when you've got something up and running, it's got to be tricky. It's just easy replacement because you basically bring YBox but all the operation model stays the same. Then the next stage is to look into that's where your workload started to work. Now you have YBox in your network. You start to look into the workload of your network and then determine, do you want to use scale up solution or do you want to use scale out architecture? A lot of our customers are migrating from scale up to scale out architecture because it's easier for them to expand and it doesn't have the physical limitation. So that's the second step. The third step is that as soon as you determine your architecture, you will start to realize you have to manage the network. Now it's becoming more complicated. So you have to figure out a way to automate the configuration. And these are the things that we help our customer by bringing Linux as our platform into our platform. People can bring the tools to program the platform. And then the fourth step is that once you have this architecture management tool YBox in the place, you can start to bring in the new software or new SDN technology to work with your system. With that, we have a cross flow. We have a certain technology that can blame the SDN open flow architecture with the current architecture. So we basically help us, our customers to go through four steps. First step, YBox. Second step, architecture change. The third step, automate all the configuration and the fourth step to reach the SDN technology. And how long does that take generally, or some examples? It depends on different customers. Some of the customers can go from one to four within one year. That's really quickly. But some of the customers are more conservative. They want to make sure they have every step proven. Every step is no downtime at all. So some of them may take even 18 months, two years to carry out the whole transition. So it really depends on customers and depends on the size of the network. Some of the other customers basically just bring up the SDN into their network. They got a sizable network, but they determined to go to SDN within six months that everything is running. So it really depends on how you want to migrate your platform all the way. And the key issue people try to solve is not CAPEX. I think CAPEX is the important issue they want to solve in day one, but the real issue they want to solve is APEX. The network is becoming more and more important and more and more complicated. If you don't bring in the new technology software to manage the network, you have to continue to hire a lot of people to manage a small size of network. And when you try to scale out or scale up to scale your data center or scale your network, you run into problem of human. Can there be multiple OS's in the data center and how big is the market for white box OS's? I think there definitely can be multiple OS for data center. It depends on how you want to use your system. It's just like everybody is believing that Microsoft is dead, right? Redhead or Linux is taking over the world. But if you look at Microsoft, they still have a big chunk of the market. So definitely we can- And they're using that in the cloud game too, with Azure. You're seeing Windows the sequel on Windows. Exactly. So I think there's definitely a room for multiple OS vendors in the networking sector. How big is the white box market? I think most of the big data centers, especially people that come to OCP, you can talk to them. You will start to realize people start to use white box in the data centers already. Even though it's not really well known, not a lot of people come out to talk about it, but white box is penetrating the market. Cisco's revenue is 30 million, right? So just think about white box. If white box grow to 10% within a couple years, that's three billion by itself. That's a huge market. And by bringing more and more people to come to the market and can customize their platform or customize their solution, you've got more and more networking functions migrate over to white box. So my gut feeling right now is that even though not a lot of people coming out to say we completely changed the network, there are a lot of vendors or a lot of customers who are testing white box already. Another proof is that even small companies like PICAI, we have more than 350 customers worldwide. So a lot of people are using white box in their production line. A lot of people are using white box in their test lab already. Talk about your business, okay? Where you guys are, the founding, the story, and what are you guys doing solving your customer problems and what are those problems? Right. And where do you guys win and where are you guys improving? What? So PICAI started in 2009. That was long before the white box idea came up. The open flow kind of market. It was started with the open flow as a penetration point, but in the very beginning, our vision was very simple. It came along with my background. I worked in tandem for about 10 years. And before I come out to start up, I was going through the whole change of mainframe to PC. So I was on the mainframe side watching everything from the down. And that was an interesting experience. Fault tolerance, all those systems, banking, trading. Exactly. We owned about 98% of the banking market. So at that time, we said there's no way. So high availability was your DNA. Exactly. Fault tolerance. Total innovators. Exactly. And we tell people, hey, if you need a huge telecom system, you need tandem database, right? You need a banking system, you need tandem database. Everything is a tandem database. So we look at things through our lens. Everything is a database application. What we didn't realize is that when you go to PC, you don't need the database at all. People try to solve different problems. So in 2009, before I started this company, we're looking at the market. I got a feeling this is happening again to networking. Actually in 2000, the same thing happened to storage. NetApp and EMC came up, right? And in 2006, 2007, Android is starting to come out and people start to realize, well, there's yet another operating system. At that time, I started to watch what chip company, what ASIC company is doing, and I started to realize ASIC from the off the shelf commercial Silicon is catching up with the big companies like Cisco's internal chips. So if that happens then- Good enough, it comes good enough, right? Exactly. Then you're going to start to see the whole market is going to migrate over to these Y-Box vendors. And we saw that coming in 2009, we started saying that, what do people need when Y-Box is coming? An easy answer is that while we create a software to do layer two, layer three. But that doesn't solve the problem because every box is small, you are going to end up managing a lot of boxes on the- So let's take a step back because one of the things we watch at Silicon Angle and the Cube and Wikibon is networking, you know, Stu Miniman's all over it, Dave Vellante, David Floyer. You know, we're totally watching the landscape. It's not just at the lower level open compute at the cloud game, it's the battleground is in the network layer. You look at VMware, Google. Exactly. You can just read the tea leaves, it's pretty obvious. That's what the battleground is. Past layer, there'll be multiple middleware layers. Okay, buy that, little cloud found you over here over there, whatever. No problem. Network layer. Why is the battleground so hot in network and what are the innovations that have come out of SDN? Just going back since this year was bought by VMware. So if you look into IT infrastructure, there are basically three components. Computing, storage and networking, right? Computing and storage have been virtualized and these cloud, that enables the cloud. However, in the past people basically just covered the eyes and said, hey, networking, as long as we have enough bandwidth, let's just keep it at the corner and say, we don't care. Let's figure out how to solve everything with the software. But as you grow your data center, you start to realize networking is going to be your limit. In fact, there are a lot of talk about barriers, right? The scalability barrier, the flexibility barrier, deployment barrier, it's all about networking. So people start to realize that, hey, we solve a lot of problem on the storage and computing side, but because of networking, we cannot deploy anything. So now all of the engine is down to networking. And this is 25 years problem. This is not just yesterday's problem. So it takes time to solve the problem. And it takes time to bring new vendors to the market, bring new supply chain to the market. What's the bottleneck? Is it the technology? Is it hurting the cats on the vendors? Is it standards bodies? Is it- I think the real bottleneck is the usage model. People have this mindset that, hey, networking is something I shouldn't touch, right? So if I want to touch the network, I bring the Cisco guys sitting by me before I make the change. So that's- Because of disruptive operations. Exactly. There's an impact. Right, reliability is always the key of the networking. However, the complexity and all the slowness to deploy things is really eating customers right now. So I think the most important thing now is that to start to realize, networking is something that you have to change. And that mindset is the biggest barrier. Once customers get over the mindset, we got a lot of tools out there. We got YBox, we got OpenFlow, we got SDN, we got Programmability, DevOps. All these things can help people to migrate to new networking. So if you- What's the pain point on that? Is it heavy duty, migration? What gives a taste of like, I'm a customer like, hey, you know what? I got DevOps in my future. I got, I'm hiring a lot of app developers. I got containers being deployed. I want a programmable infrastructure and I want a data center operating system. Come on, what do you got? Right. The pain point is not really the technology by itself. The pain point is the business model. So people try to solve the pain point that, for example, I'm running a container right now. Cisco is telling me to do it this way. Now I need to add another container. What do I do? I have to hire another guy. So you have to deal with those problems. And by dealing with those problems, you need to figure out an open way to program your network. You need the tools to bring the hardware. You bring the tools to bring the new application. So these are the pain points that people have to deal with. So you said that people didn't want to touch the network. That was, don't break it, it's fine, just add more bandwidth. Once they've made the lead to SDN and they see the value, are you finding customers really see the value? Not beyond just going, getting away from the barrier, but really opening up new opportunities to do things that they could have never done before. Absolutely. I don't know if you can share any good stories with that. We have a customer that runs three data centers. Inside the data center, every time they create a new data center, they have to deal with a couple of problems. You have to put the infrastructure in place. You have to provision your network. You have to bring network engineer. And because the data centers are spreading over the US, you now have to set up the team in US over the global team. And they set up the policy before they develop their SDN solution. They set up the policy, I want a zero touch network because for server, for storage, I can do zero touch, no problem. But networking, I always have to have somebody on site to program the network, to watch the network. So they set up this policy that they want to do zero touch provisioning, zero touch, zero touch, fault tolerance, that kind of things. And you need the software to do that. So once they, thank you. Once they solve that problem, they now are running three data centers. They are looking into the next three data centers they want to start. Guess what? In the past, it takes six months to bring out three data centers. Now it takes days. Okay, yeah. And that's the solution. Once you get through that, from three to six, six to 12, you just don't have any barrier. You just continue to roll out your data centers. That's great. So the question about the market is one final question for you. What do you guys want to do? What's your objective this year? I'll see you guys have been out there. SDN is super hot. Internet of Things is exploding as an application, which essentially means mobile computing and every other connected IP device. Exactly. What's the future for you guys? We want to change the network industry. So that's our mission. Our mission is to change the network industry from a closed environment to an open environment. We cannot solve all the problems, so we have to build up a lot of partnership with our partners to solve the hardware, supply chain problem, to solve the chipset problem, to solve the application problem. But essentially what we're trying to do is that we see software defined data center as the end game. You want everything to be virtualized so that today, if I'm running virtual machine in the US, tomorrow I can run it in Australia. And the day after, if I'm traveling to Japan, they can follow me to Japan, right? So this is the vision. The vision is that we want to enable software defined data center. And in order to solve that problem, you have to solve the networking problem. In order to solve the networking problem, PKI has to be the one, or at least one of the vendors to enable the change. Right now, if you don't have a way to segregate the hardware from the software, virtualized data center to physical data center, or the control plane from the data plane, you cannot get this to happen. So what we try to do is to focus on how to disaggregate those areas and enable other vendors or other partners to solve the problems. James, Leo, thanks for coming on theCUBE. And remember, when you're a billion dollar company, don't forget us, okay? Because, you know, you were once on theCUBE. We're doing our part going out to the stories. Thank you so much. We have a great opportunity. You guys are in probably one of the most important areas in the technology industry right now. This is exciting. There's unicorns in this room. You guys could be one of them. You've got a great product, great team, and it's hot areas, so best solution will win. It's not marketing. You can't market your way in this market. We're lucky, right. Congratulations. It's theCUBE live in Silicon Valley. We'll be right back with our next guest after this short break. Open Compute Project Summit here in Silicon Valley. This is theCUBE. I'm John Furrier with Jeffery. We'll be right back.