 on the ground, presented by theCUBE. Here's your host, Jeff Frick. Hey, welcome back everybody. Jeff Frick here, we are on the ground at the Santa Clara Convention Center at the Open Networking Summit 2016. It's kind of the season of open shows. We were at Open Compute Project Summit last week. And again, wanted to come down, kind of see what's going on and really excited to talk to Calvin Chive from Pica 8, welcome Calvin. Thanks Jeff, good to be here. Absolutely. So like I said, we were just at OCP, talked to James from Pica 8. I kind of cheated, I got a little update from him just last week. And now you guys are here at O&S. For the folks at home, what's the difference between OCP, O&S? How are the two related? Why is Pica 8 at both? We'll go with the open, right? So that's the season of open shows. So I mean, that's a great question. Open Compute is really about, for us anyway, you heard the talks last week, it was all about the interface of the hardware, right? How are we submitting designs from an open source standpoint for the hardware? And from Pica 8 standpoint, we've done a lot of work in that regard, like how do you port to ASICs faster? How do you make memory space allocations better with ASICs? So that's all really important for us. The other aspect of that is, how do you interface with the apps? So for us, open networking is all about the apps. And if you think about it, it's about how do you customize the network for the apps and not the other way around, right? If every app was the same, then this would be a really easy problem to solve because you'd have the same application characteristics. You'd be able to build the exact same networks with you just be able to go down to Cisco, Juniper, Arista, whatever your networking vendor of choice, pull down these boxes and build the network that you need. But the reality is that with all the application customization that's happening with all the diversity in the types of apps that we see out there, it's a lot harder to do that. So you need to have some custom networking components as well. It's interesting, open source for hardware versus open source for software, which everybody's familiar with and actually the Linux Foundation now has taken over open networking some bit. But in hardware, it's a little different, it's a little different animal because you're trying to come up with a set of standards that people can build to, but you still have to have competitive differentiation. So how do you kind of straddle that challenge? Yeah, that is different. And there isn't, I mean, it's probably an oversimplification how we make that demarcation line. But for us, it actually works out really well because we actually sit right on that middle part, right? So we build the hardware operating system, Picos, that sits on that open compute standard hardware. And we also have the interfaces, we also support the protocols that you need to communicate with that application ecosystem. So when we think about the applications, it's not just the app, it's the SDN controllers that you have to think about, it's the automation platforms, it's the management platforms. So there's a whole slew of these different components that you need to talk to, and that's really where the interfaces come in that we support. It's interesting, there's always conversation, the three basic components of compute, store compute and networking, and that networking was really kind of the last one and SDN coming around, but software is finally getting into networking and those kind of all-in-one boxes are being broken up into pieces. And as you're saying, it's really a function of applications and being able to morph to this rapidly developing app world with DevOps, et cetera. That's right, Jeff, you've actually just described the network gap and we call it that because if you think about how new applications and services come online, if you have an application developer or programmer, you can actually spin up an application pretty quickly. You just write some software, you compile it, roll it out, off you go. To make a networking change of the same magnitude takes a lot longer cycle, at least traditionally. And so there's this gap that exists because you have to wait for a new ASIC to be run off the fabrication line or you'd have to wait for a new version of the operating system code. But now that that's all been disaggregated, you can actually close that gap and furthermore with SDN protocols, you can actually make the communication between the application and the networking component a lot more tight as well. So that's really helpful to close that gap. James had a great line and you expanded on it since we've been here about before you would have to make everything fit your network. You didn't have the flexibility of the network, so your storage had to fit the network and you took it up a notch. It's really now you can make your network fit your application versus making your application fit your network. Pretty important in the world in which we live with DevOps, fast moving application space, API based economy, really a huge enabler. Yeah, and that's exactly what you need to do, right? Because if you, you know, one thing that if you've walked around on the show floor is one thing that you've noticed is that the degree of customization required for these applications is pretty intense, right? If you, you know, one of the themes of the show that I've noticed so far this year is that we've moved from science projects and proof of concepts to real production, but there aren't cookie cutters. I mentioned this before in terms of everybody's application is different, right? So how do you build the infrastructure to support that? You really need to have, you know, malleable, dynamic, you know, open networking components that you can actually like change on the fly. And as an example of that, we've got a demonstration with a company called CoolCloud that's actually downstairs. And what we're doing is it's a really simple example. It's a cloud bursting example where, you know, when you're thinking about provisioning capabilities and resources on demand, you really need to be able to call on that quickly, right? So if there's like some kind of a congestion event for instance, and you need to spin up another instance of that application, how do you do that quickly? In this particular demo, what we're doing is we're using the Picos with an open flow controller with a CoolCloud software, spin up another virtualized instance from somewhere totally across in another network. It's not even next to it. And then you can actually get that bandwidth that you need. And when you're done, you turn it off, right? So it's a very simple cloud value proposition, but we're starting to see the reality of these types of use cases come to life using SDN and standard networking. It's a simple value proposition, but it is at the crux of the cloud revolution, right? It's really hard to do with standard network. Absolutely, because it's all about having capacity on demand. So whether you can plan it, like you're running a Pepsi commercial in the Superbowl, right, the classic example, or it just got, you had the other Superbowl that had a lot more traffic than they anticipated a couple years back. And that's the thing, right? Is because you can't always predict because what the provider wants to provide the consumer is you dictate when you want that demand, right? It's not I'm provisioning it for you, but you want to be able to turn that up from a consumer standpoint. So customer, you can't predict when their usage patterns are going to have to happen all the time. So you need to be prepared for that. And so that's really kind of the interesting thing. One other thing I want to add really quick is that, you know, one of the biggest misconceptions about open networking and SDN is that you have to like rip apart your entire old networking infrastructure to make this new model work. That's not true at all. In fact, you know, if you think about it, there's been 25 years in the networking industry. These routing protocols are really mature. They're very robust and stable. These things work really well. What we're advocating is think about how to integrate SDN into those environments so that if you do have a specific problem that's a problem that's difficult to solve, whether it's dynamic VPN provisioning, whether it's cloud bursting or something like that, that's where you bring in the policy that you can make with SDN. Well, let's unpack that a little bit deeper. So are you saying if you've got kind of an existing infrastructure based on stuff you've had for a while and you want to start to add this capability, you can basically put your stuff as kind of an overlay layer to then draw off those existing resources or is it like a separate pool that's then available on the more flexible basis? Yeah, well, that's a great question. That's the thing is because a lot of folks won't want to throw away their investment, right? It's time consuming. There's a lot of optics involved. Plus there's so much institutional knowledge that's built up over time. It's like, hey, I know how BGP works. I know how MPLS networks work. No, you absolutely don't throw that away because that stuff works for what they're being used for. In certain cases though, if you want to do, for example, fast multicast convergence, if you're trying to spin up a services gateway, if you're trying to roll services out to the edge, rolling fiber to the home, for instance, these things are sometimes easier to be done on a fast basis using SDN protocols in conjunction with those things. So it's not like you're turning them off. You're actually doing exception based forwarding as opposed to bulk based forwarding. Okay, so you've been in the business for a little while. Before you were at PK, you've been in networking. So what's kind of next? In the next six months, nine months, 12 months, what's kind of the next mountain that you guys are trying to climb? Yeah, no, it's really interesting. From last year to this year, we're already starting to see a lot more production deployments. I think I mentioned that before. So a lot of these cloud service providers, whether it's infrastructure or content providers and even some large enterprises are rolling these things out. The customization is key. So for us, it's going to be to continue to work with partners. It's going to continue to work with OEMs. It's going to be to continue to work with existing networks out there to how we can add value to those existing networks using our operating system, using open networking concepts, using OpenFlow. Okay, I was going to let you go, but now you just made me think of one more question. The ecosystem, right? It's a very different model when you're leveraging an ecosystem model. And we get to a lot of shows and most of the shows we go to are shows because the supporting company has a large and vibrant ecosystem. How is an ecosystem involvement really changing the network game versus kind of a closed system you bought either box A or box B? Yeah, well I think the interesting thing is open source, right? That's always been one of the things is it's a little bit of a double-edged sword because you release something to the open source ecosystem and there's actually a lot of communities out there right now. And I think there's probably a little bit of confusion in the marketplace as to how all these things fit together. So I think that'll be the biggest challenge. One of the bigger challenges moving forward is to kind of really understand where all these ecosystem players play. And usually when you open source something, like the idea is sound, but it takes some iterations before you come up with the real, how should I say this, like the robust or like this is the right solution, right? It's almost like here's the design, now you 10 groups go out and figure out like how to build it and there's 10 different implementations and over time that process evolves until you get like a really nice one. And I think that's what's going to happen here in the open networking space. Okay, yeah. It's always that trade-off between speed of innovation with a lot of people contributing then, but you can't have too many projects for the commercial customers, right? They're like, ah, what am I supposed to do? To find a group of developers, right? Absolutely. All right, Calvin Chai, thanks for stopping by and sharing some of your insight. Yeah, thanks, Jeff. Absolutely. All right, Jeff Rick with Calvin Chai from PK. We are live at the Open Networking Summit in Santa Clara, California. Thanks for watching.