 Okay, we're back here live in Las Vegas with HP Discover. This is SiliconANGLE's exclusive coverage of HP Discover 2013, SiliconANGLE. This is theCUBE, our flagship program. We go out to the events. Extract a signal from the noise. I'm John Furrier, the founder of SiliconANGLE. I'm joined by co-host Jeff Frick. And our next guest is Dave Larson, VP and CTO of HP Networking. Welcome to theCUBE. Thank you, good to be here. Good to see you sporting the tie today. Very corporate-like for a tech guy. But let's talk about networking because obviously the convergence equation, software-led infrastructure, as we call it, Wikibon, our analyst group, is really networking servers and storage. So Dave Donatelli talked about software-defined storage. NetApp had a big announcement that software-defined fill-in-the-blank these days, right? Yes, it is. Very hyped up, but that's an indicator of a sea change that's coming. Correct. OpenFlow started this. We saw software-defined networking, which is basically network virtualization, started all. And then the Siri got popped out by VMware and then all hell breaks loose. Yes. It's a whole new category, right? That's kind of a read-us-ditis version of what happened the past two years. So what's your take on all this? You guys are the first ones to ship OpenFlow. You have a lot of ports that you've sold. Networking's in that equation of those things that are now integrating together across diverse environments. Give us your speed and your take on that. Yeah, I would say software-defined networking is offering a new opportunity to rethink the way we build our networks. Topology-wise, policy-wise, how to extract business value from that critical investment that today is really viewed by most CIOs as a blocking mechanism to achieving the things that they want to on a day-to-day basis. So SDN is kind of a very timely acronym, but a lot of people are using it in ways that are not necessarily correct. So an example would be, for instance, would be the actual one that you mentioned, NYSERA being acquired by VMware. It's interesting technology. It's very important technology. Network virtualization is critical to allowing hyperscale cloud environments to be able to deliver the orchestration automation that they need to meet their customer demands. The problem is that that in and of itself is not strictly SDN. It's a great SDN application, and we are building a very similar SDN application with HP Cloud Services and HP Converge Cloud to scale in similar ways, but that's not the end of the story with SDN. SDN needs to matter to enterprises end-to-end from their branch through their campus, their campus data center out into the cloud through the WAN, so if you don't provide the kind of policy controls to enable real value there, then you're just a one trick pony. So why isn't NYSERA, I mean, they've kind of dodged a little bit. I mean, so I've interviewed Martin a few times on the queue at VMworld. He's great to have on the queue, he's great dynamic, but when he's on the queue, people comment on Twitter, and so the crowd is watching, right? You're live, so one of the things that he gets dinged on is that he, oh, I'm not even worth a flip-flop, change directions, and then I asked him about that, he said, the market has changed. Is that your take on it? I mean, open flow to NYSERA, to what we have here today, what's going on with SDN right now? What is the SDN? How do you package it? How do you explain that now? Well, the market is changing. I mean, this is evolutionary. We're at the beginning, we're at the advent of this meaningful paradigm shift in the networking space called SDN, but it's not clear what this is going to look like in five years from now, although I will tell you, it is clear to me that it is not just going to be strictly a data center technology, it's going to be a technology that unlocks value for enterprises at the Ethernet access layer, at their WAN DMARC, at their load balancing requirements that they have for specific applications, and it's more about enabling the applications themselves to request policy that'll allow them to behave properly than it is strictly about virtualizing the network, although it is partly about virtualizing the network. Let's talk about networks, right? So I was talking, I wasn't accused, I was off with someone else about the Chinese taking the packets in the US, basically forcing the routers, the least cost pass, and taking the packets, and that was just an example of security. This is a casual conversation, but cloud providers, service providers, are running networks, and they need to integrate into the enterprise. There's not, again, I would agree with you, it's not just the enterprise. How does someone, I mean, how do people make sense of that? I mean, you're in the weeds technically, but also you have a landscape view of the marketplace. You've got enterprises and you've got external networks. What is going to be those ways to route traffic? How do you run virtualization across different networks? Is this going to be an answer there? Is it still to be figured out? What's your take on that? Well, I think it's, you need a policy control mechanism that's going to allow you to take advantage of the elasticity and the economics of cloud, right? There's a lot of people that are saying, hey, you know what, I want to take advantage of cloud because it looks to me that it'd be cheaper to operate my environment in the cloud. And yet I have security policies where my business is highly regulated. How do I do that? How do I get these applications into that environment? Well, if you don't have a rock solid policy control point that can interface your branches and your campus environment and dynamically allocate resources in the cloud such that you always have control and you can attest to that. Well, let me drill on that. Because we were talking, Dave Vellante is not who will be coming here tonight. Dave, we miss you, we'll see you tomorrow. Jeff's doing a good job. He's going to get a question, I promise. Policy, static-based policy is dying, right? People want dynamic policy. So, you know, on the storage we're seeing it. I mean, you've got manual policies or even policy-based access setting up pre-policies is that's not the trend. The trend is I want a machine learning. I want learning machines. I want dynamic policy. Virtualization is a nice way to do that in software. So how do you see that trend? Because policy is great, but when you have all kinds of different policy opportunities, it's hard to set that statically. So what's your take on that? Well, let me use a specific example in answering that question. One of the problems that we have in enabling or enforcing policy across a network environment is that many of the applications that we're utilizing are encrypted today for a variety of reasons for security and for control of data. A good example of that is telephony and collaboration technologies like Microsoft Link. So we built a proof of concept technology in our SDN switches using our controller that enable tight integration with Microsoft Link server to enable real-time dynamic QoS policy in a network without any kind of network administrator involved. And the way we do it is we actually use APIs from our controller to interface with the Microsoft Link server so that when the endpoints and a Microsoft collaboration environment using Link decide to open a session, we can identify it and then we can provision the QoS and the links associated with that on the network in real-time. All right, we got the sounds like an intro to it. That's a damn beef going on over here. Sounds like a porn movie to me, but, you know. Anyway, I wouldn't know. Actually, it's, I'm so all updating myself. That's 70s. Okay. Oh, the cube is unedited. We're alive, doesn't matter. So we're talking about dynamic policy. Let's take that to another level. Okay, so what does that mean for virtualization? Because the hope of network virtualization was to have the software being programmable. So how does that work in and how does big data fit into that? So big data is a similar thing. Big data, very similar to Link, is going to require on the nodes that are cooperating on the computer or whatever they're doing for the big data operation. They're going to require certain policy on each of the links, and maybe in a very dynamic. For me and Chad. The music. There's driving everything on the next team up. Are we going with the sound? Okay, now we're good at the sound. It's kind of distracting. From a big data standpoint, the policy still needs to be instantiated in real time, based on whatever is going on in the application. You can expect a network administrator to be monitoring and making changes that would optimize the application itself. So from a virtualization perspective, the ability to use a controller interfaced in an open and standard way with the application to then provision the network to best meet the needs of the application is the policy at the end. So we're going to hear the background here. Meg Whitman is about to take the stage in five minutes. Dave, I got to ask Dave Larson here inside the QBHP Networking CTO. Now at Apple's World Wide Developer Conference, the big buzz, one of the big balls. One of the big buzz we noticed we get into was the 12 core Xeom. That speaks to some of the compute that you guys are involved in a lot of also, the science of also the chip set, it's not just software, a lot of actually awesome compute to go along with some of that software. How does that all play into it? You mentioned some dynamic, that means you can go into the database, maybe some flash, the stuff that the storage guys announced they'd pretty compelling. So you have to keep an eye on that. So can you tie that together? How does that all hang together? So as Dave Donatelli says all the time, we're about delivering a software-defined data center. And that is built on top of our Converge Infrastructure Platforms. Those Converge Infrastructure Platforms, everything that we do is tied together by the network. So underlying that is HP Networking, software-defined networking technology that allows you to best take advantage of those environments and actually bring the policy and the control from the storage side of the network all the way through to the end, end station that's consuming that data in a very dynamic fashion. Well we keep you out all the time. I know Jeff wants to get some questions up before one last question is, we keep you out on that data center conversation because that's interesting. And we've talked to Dave about this, about the operating system in the data center. If you look at the data center as an operating system, you have so much big data within the operating system, probes, circuitry, policy data. So you can access all this input coming in, all these decisions going on in real time. Dynamic, as you mentioned, how does that all go together? What has to happen on the software side up the stack? So just take us up, what are you looking at as a CTO in the networking guys? What are you looking up at the stack? What do you have a platform layer? What do you see up top above you? What kind of software? Up above the network? So we see applications, right? We look at the SDN control or as the operating system that is provisioning the underlying network from a policy perspective. Above that, we talk about SDN applications. An SDN application might be load balancing. It could be a QOS application like I talked about, Link. It could be a security application like we talk about with our Sentinel security that delivers IP reputation at the ethernet access layer. All of these things are tools that allow you to abstract the capabilities of the data plane. And then above that are the applications that run your business. The key premise is that the applications that run your business need to be able to have a mechanism to communicate with that control layer, my operating system, such that we deliver the policy that is meaningful to that application and can deliver the needs of it so that you don't have congestion, so you don't have applications timing up, so that you can actually differentiate the needs of the business and the real business value. So let's get technical on this breakdown. So Dave Donatelli says that means we got to break down the silos. This box does this, this box does that. I need this API for that. So that means you're basically saying I need a control layer. And you need one fabric. You need one fabric that is connecting your storage, your server, all of your applications and delivering it throughout your... What fabric is that? Do you think that fabric? Yeah, it's HP networking that is delivered and it's not all physical, right? Some of it is virtual. We have a virtual switch that lives in the hypervisor and delivers functionality in that environment and exports it out into the network control, the network data plane where ASICs can actually provide capability. This is the magic question. So HP fabric is the platform. Is SDNA mindset or an architecture right now in today's marketplace? SDNA is a tool. Software to find networking is a tool and it's an emerging paradigm for breaking the location of the data plane, the control plane. In and of itself, it's of little value. If you don't deliver an application that's utilizing it that extracts the value from the network, then it's of no value to it. My question is, what does that enable value-wise on the application side? That's not being able to be done today. A good example would be our Sentinel security application. Today, enterprises never walk down the Ethernet access layer because it's too costly. If you think about a wiring closet that has a number of gigabit ports in it, maybe hundreds of gigabit ports, are you going to really choke that through a one gig security appliance? No, you're not going to do it. You're going to allow that environment to access the rest of you. Now we're going to steer flows of interest as they get closer to your data center through a security layer. But that's not sufficient. Enterprises are screaming out for that technology. They're saying, oh, I'm getting infected. My end users have a BYOD environment. They're taking their laptops home. They're getting infected. They're bringing them back and then they're propagating malware in my environment. How do I fix that? Well, SDN allowed us to say, well, wait a second. We have some technologies that will allow us visibility into that, namely our IP reputation called RepTV database. That's part of the tipping point. Why can't we instantiate that at the first hop in the network? Utilize our controller technology to intercept all DNS lookups, all new IP address lookups, and determine whether they're trying to access botnet command and control. As soon as you know how to do that, you stop the network right there at that port. You allow no propagation of an infected machine. We have customers that are using this proof of concept technology in live environments because it works so well. So I got to ask you about one of the incumbents out there, one of your competitors in, or someone competitor, could be competitors. Yeah, in this case, they direct competitors, Cisco. Enterprises have a lot of Cisco gear. Routers, switches, balancers, you name it. They got it all. You guys have made some good strides against Cisco. HP's got 3Com, and HP's networking has just had a great track record over the years. Okay, that's great. How do you compete with them? They're putting together a big SDN group. We're in the street and they've got some cobbling to get us some stuff. So how do you guys compete in those Cisco environments with HP networking? Were they're trying to trip you guys up on the SDN side because they don't have gotten there yet? We can be with them on open standards, right? At the end of the day, everything we could do is going to be based on open standards. So, open flow. We're a huge advocate of using open flow in the switches and in fact, all of these applications that I've been talking about are delivered on our switches utilizing open flow, of which we have upwards of 40 million ports in deployment that are open flow enabled in the world today. Cisco wants you to use their proprietary 1PK API. And then program the infrastructure. Explain for the folks out there and our vloggers who are watching might be vlogging this up. The interesting, why is open flow SDN or not vice versa? I mean, I consider it all one part of the tooling. Why, what is open flow? This nuance is there. Open flow is the only available standard protocol with which you can instantiate SDN on a switch today. Everything else is proprietary. Open flow is SDN, not vice versa. Open flow is a component of SDN. SDN is the abstraction of the data plane from the control plane. And open flow is the standard space protocol that allows you to do that. And if you're using open flow, you are doing SDN. Yeah. There it is. If you're using open flow, you're using SDN. Great, you want to just get that out there. Because a lot of people are looking, there's a lot of fud out there right now on open flow, this version, that version. I see it as obviously not about a VM ware. Well, we do SDN, that's not open flow as well. So we do SDN with open stack. So open stack, again, a standard space initiative within the cloud to enable virtualized networking capabilities in server and application environments. So our controller also contemplates how to manage open stack in a hyperscale cloud environment. And the marriage of those two is the only standard way. So will you add open stack summit this past? I was not at open stack. I was at open networking summit, which was concurrently held. Yeah, a lot of guys, the big switch guys were running back and forth too. So we did three days of live coverage here at theCUBE and it was really awesome event. So can you just share with the folks out there some of the contributions you guys done to open stack? Yeah, we've been one of the early contributors to open stack. We are one of the most significant code contributors to open stack. We base our cloud environment, our cloud services is based on all the stack environment and we have integrated it in every aspect of our SDN solution. Awesome. Dave Larson here inside the CUBE HP networking. We've had Mike Bannick on the past. HP networking is just a steady train that keeps plowing down the tracks. You guys doing a good job. And again, Donna Telley's got the magic kingdom there with storage, servers and networking all coming together and software led. So congratulations, software led, software defined, Bill in the blank, it's here. Mark Andreessen said software's eating the world. This is theCUBE. We'll be right back with our next guest after this short break. Thank you, John. Thank you, Joe.