 Okay, we're back live here at the OpenStack Summit in Portland, Oregon. I'm John Furrier, the founder of SiliconANGLE.com. I'm joined by my co-host, Dave Vellante from wikibond.org. This is SiliconANGLE's theCUBE, our flagship program. We go out to the events and extract the signal from the noise, and certainly here at OpenStack, there's not a lot of noise, but a lot of signal, a lot of developers, a lot of use cases. Really, really, the alpha geeks, the practitioners, really putting new technology into place to power this modern era of computing, cloud, mobile, and social. David Floyd, we're here with Dimitri Stiliatis from Nudge Networks and Mountain View. Welcome to theCUBE. Thank you. David, I want to get your take on this before we set up this interview, because obviously we've heard from RightScale, they're on the management side just previous. We've had Rackspace on earlier. They're on the provider side. We had Big Switch on, software defined networking. And now, Dimitri's company. The software is eating the world. What's your take on the SDN market right now? Relative to OpenStack. Relative to OpenStack? Well, what you're clearly wanting to do in every part of it is separate out all of the different layers. And you ought to be able to separate out the physical and the logical and the software is the way that that's going to be done. So instead of having to have a switch, which is a piece of hardware and the software, you want to separate the two out so that you have the logical function and the physical function from the two pieces. So that's very important to be able to contribute to every layer, take new technologies along with you, and then define the software element of that as the piece that you keep constant as the technologies themselves adjust. So durable code, but yet manageable, you can build on and maintain. Which can take advantage of new technologies as they come along. And obviously, coming back to you, what are you contributing? What do you think needs to be contributing? What's the white space in that area that you're going after? So, see, when people started thinking about the cloud and OpenStack and all this kind of thing, they quickly realized that the network is a fundamental piece, right? You have to start with the network. You have to interconnect your components and so on. The angle that we are taking is, yes, it's good. Within your data center, within your cloud, you have to create these network services, interconnect applications and so on. But much more importantly, you need to be able to dynamically connect these applications with your existing network services, right? So you have a large amount of enterprise VPN services. You have hybrid clouds coming out. So you need to be able, the moment you activate a network service in the data center, to be able to seamlessly interconnect this now with your enterprise sites, with other network services and other data centers and other clouds and so on, right? So the network is always a network of networks and we have to bring everything together. We cannot just restrict ourselves within the confinements of a single administrative model. So that's a fundamental part of what we are trying to bring here together. Okay, and so how are you fitting in with the network layer? So our view is that, first of all, we need to talk both languages. If you don't think of it as a translation thing, right? So we need to understand the language of the cloud. We need to understand the language of the application developers and the cloud. They want to use some abstract mechanics to define their network services and install them if you want in the hypervisors and open stack quantum seems to be the prevalent way to do that. So that's language number one. But then we have all these thousands of networks out there where their language is BGP. So what we are doing is we are mirroring the two. We allow you to code and define services in open stack and we allow you to define the mechanism in the interconnect the services automatically with all the other networks that are out there. So I call it, sometimes we're just translating between languages. All right, a language translator. From an application point of view, they want to consume resources and previously networks and computers were the main things they consumed. But it seems now that, sorry, computing and storage were the main things they consumed. But it now seems that networks themselves have to pay a much bigger role in providing a quality of service to those places. And you've got a quality of service down in the nanoseconds when you get to the server level and used to have milliseconds for the storage side. It's now coming down to microseconds. What are you doing to make sure that that quality of service is not just the bandwidth but it's also the latency? How are you planning to manage that? See, the way data center networks evolve is people are quickly realizing that the same, if you want principles that we used in order to build the internet itself can be used inside the data center. So if you think about the internet, right? In the internet, there is voice services, there is video services, there is all these other services running. And they are actually running by assuming you have a well-engineered IP network and then you run the services at the edges if you push all the intelligence at the edges. It's the same thing where the network on the data center is going. The data center network becomes a very scalable IP fabric. It is very well managed if you want, very well trafficked and engineered. And you push the edges at the hypervisors. You push essentially the services at the hypervisors where traffic is differentiated. So if you see, for example, a tenant misbehaving, you're gonna block him at the hypervisor layer. If you're gonna provide QoS or map different tenants to different classes of traffic, it's happening at the hypervisor. So the center of the network behaves like a scalable IP fabric and all the intelligence it's pushed around the edges. And the reason you want to do that is because this allows you the ultimate scalability, right? The network core doesn't need to know about every flow that goes through the core of the network there, right? You don't need to know the IP addresses of virtual machines. You don't need to know what individual virtual machines need to know, want to do there. You just need to worry about aggregates so you can engineer and scale the core, make it very cheap. And because you make it very cheap, you can increase the capacity of the core and you can distribute all the intelligence at the edges of the network. Right. So you said that you can do to the hypervisor and that's obviously on the compute side, that side of it. But what about the data network? Isn't that a, don't you need to regulate the priorities and flow from the data through? And isn't that today, that's a very big part of it, isn't it? Yes, but it is still happening at the hypervisor, right? The first touch of an application with a network is not anymore the top of rock sheets, let's say on the data center, but it is actually the hypervisor virtual sheets, right? That's the first time that you see a packet. When a packet comes out of a virtual machine, the first time you see it is at the hypervisor itself. And at this layer, when the first time you see the packet at the hypervisor itself is where you apply all your policies, right? In other words, the edge of the network is not the hardware, it's not the sheets on the top of the rack. The edge of the network is inside the server now. Okay, yeah, okay, excellent. So I want to ask you, we have a couple minutes left here. We have two minutes left, I want to get your perspective on the state of the business around OpenStack. What is your view, okay, because you're chief architect, you're looking at the tech, but you have to intersect the business objectives. What are you seeing as the core business drivers that are causing you to make your technology a certain way? Right, so it's clear that what people want to do is they want to provide this ability to their end users to consume services rapidly, right? That is what is driving this whole OpenStack development and more important, the community came together in order to unify if you want the core engine and the core APIs in order to make this consumption of services very easy. And in order to allow the application developers to move from one cloud to the other and so on, right? What we do is, what we try to do is, in addition, is expand if you want this model and making the network as consumable as the storage and compute facilities, right? And I'm not talking just about the network in the data center. I'm talking about also the network in the way that the service in the data center of a cloud provider will interconnect with enterprise, right? If you see the next, if you want Holy Grail that everybody's talking about is the hybrid cloud. The hybrid cloud is only possible if you can connect the network and the services in the service provider cloud with the network and services in the enterprise itself, right? So what links the two together is the network. So we have to make this network to be consumable. Final question for you is obviously DevOps is a mindset. We heard from RightScale that that adoption is in mainstream enterprises and service providers, but the word infrastructure as code is becoming more popular outside of the geeks and the architects and the coders. What, in your mind, how would you describe infrastructure as code to the folks out there? Ah! Give it a try, it's okay. No right answers. It's a moving target, that's what it is. Reality is that applications and code is a living organization. It's constantly changing and you cannot assume at any point it's static, right? It's not the good old days if you want and that's what it really means, right? It's a living organism. It will constantly adapt to the new requirements out there. Like switches in the old days, you knew exactly ports and you knew what was going on now. It's all kinds of weird stuff happening, right? It's all stuff you have to be, you have to accept change if you want, right? So it's the, actually there is an old Isaac Asimov code, right, the author of the science fiction stuff that said the only constant is change. We should do a genome project just on the network genome. Here, software defined networking, Dimitri Stilianos, thanks for jumping inside the cube. Again, you're here with a lot of the chief architects making things happen, congratulations. Thanks for joining us. Thank you. We'll be right back with more analysis from David Fleurier after the short break and a breakdown day one and day two here and more depth from the analysts here at OpenStack. This is SiliconANGLE, Kibon's exclusive coverage of OpenStack Summit, we'll be right back.