 really digging into some of the, you know, changing dynamics in the networking space. And if we talk about networking, you know, one of the biggest, you know, explosions that are going on is in the service provider space. And joining me for this segment is Luke Norris, who's the CEO of Pete Colo. So people that have watched Wikibon, actually, we had Luke on for a peer insight a couple of weeks ago. And Luke's been making the circuit. At VMworld, he did the super session with NetApp. He was at AT&T Park and he's a bit of a rockstar here when you're talking about the service provider space and some of the transformations going on in IT. So, Luke, first of all, welcome to theCUBE. Thank you and thanks for that introduction. Absolutely, you know, really, you know, love to talk about people that are really driving new technologies, you know, taking new ways of driving business and, you know, service providers in a lot of ways are helping to drive really kind of the tip of the sphere as to who's adopting new technology. A lot of ways you have, you know, less legacy than the enterprise environments and you can start with more of a clean sheet organizationally and equipment-wise. So, for people that haven't heard before, give us a little bit about your background and how you started Pete Colo. Sure, so I came from the service provider area. I spent a couple years over at Sungard Availability Services. I really sort of formed the model for Pete Colo. Pete Colo is really a service provider to service providers. We build baseline infrastructure services, allowing SaaS and past providers to build products on top of us and really also leading the transformation for the VAR community to start actually selling cloud-based services. And they do that based on top of our powered by Pete Colo brand. Okay, but you have your own facilities, correct? We do have our own facility in Denver and then we also do white-labeled services throughout data centers all across the country. We're currently in Denver and Seattle and we just announced New Jersey and Chicago coming online. Excellent, and Pete Colo's been around for how long and how many kind of generations of technology have you gone through since you launched? We've been around for six years and I think we're on our sixth iteration of the cloud. And that's actually probably what makes service providers so much more adaptable to it. Our procurement cycles are really driven by business demand instead of by three-year or five-year sort of acquisition level cycles. So we're able to see what new technologies out there address that as a product offering and bring it to the market in a much more rapid scenario. Yeah, I mean, you know, everybody is dealing with growth, but you guys have hyper-growth, you know? Could you give some specs as to, you know, what kind of growth you guys are experiencing? Yeah, I mean, in the course of a year, I think we've grown over three petabytes just within one data center, so let alone across multi-data centers, I wouldn't know that. I also know one facility's now reached about 15,000 VMs and I couldn't even go with the petabytes of RAM that that's actually running. Because what we find is actually our VMs are averaging 32 to 64 gigs of memory just within the VM itself. So I mean, you extrapolate that out over tens of thousands of VMs, it's a massive scale to deal with. We also just run into an incredible issue of VLAN overlaps and IP overlaps. You can't tell you how many enterprises still have 10.10.10 or 192.168 as their initial IP offering that they built within their data center. Even when they try to move to a cloud, they can't change that. Their applications are embedded in. How many VMs do you have in that environment? In one particular environment, well over 10,000. Okay, so one of the things that Brocade announced today is really these just giant, scalable fabric environments and they can go up to 300,000 VMs. So the question I have is, do you need that? Or I found service providers, a lot of times their networks aren't as large as even some enterprises because they have pods in different locations. So what's your experience? We actually have a design methodology and we call it failure zones and we keep our failure zones very, very small. So we wouldn't actually use maybe the generation that they just released here. Those larger, larger switches that can do hundreds if not thousands of connections. We'll actually do more discrete level switches to keep our failure zones very, very small. And can give us a little bit more granularity on that if you can, how large do we do it and is it just to keep the layer too small for virtualization reasons or what reasons? It's kind of going back to almost brocade's old methodology that they themselves built very small pods when it came to the sand-based architecture and we're sort of following through with that on this VCS and the VDX level architecture. We're building these to about 15 to 20, 60 port VDX switches and it's more doing an isolation level for how many physical servers we can connect in. We're not worried about overrunning the ARP address translation maximization, which I've talked about now. So you could see 15, maybe 5,000 to 10,000 VMs in one of these VCS pods. And then we use layer three technology to sort of map between them. No matter how stable Trill is, there's still the human error, there's still the human factor, there's upgrades that need to happen, there's generational changes. We're on the first generation of the VDX switches. We want to move to the next generation mid next year. So the ability to swap those out, you have to have those more constrained failure zones. Okay, so brocade talked about they have 700 customers now running VCS and you're one of them. What's your experience been? How did you get to VCS? Can you kind of walk us through that process? Yeah, we did a very tight vendor selection. We looked at everybody that was coming to the market and then the people that were actually at the market. Brocade was at the market. And this was probably November of 2011, maybe a little earlier. And we were one of the early, early adopters, first line of code. I literally think the code was a 1.0 code set. We went to production soon thereafter once the code set was based off. And what we were actually looking for is unifying the entire fabric. We wanted to get away from a fiber channel idea or a secondary ice guzzy fabric and actually just have a unified 10 gig fabric across the board. And brocade's expertise on the storage side really gave us sort of the comfort level that that's what we could do it on. And this was the platform to build it out of. Yeah, one of the things I found interesting about your case study is you actually, when you went to brocade, you eliminated your fiber channel environment. So one of the discussion points I was having with Mike Claco earlier is, fiber channel is doing fine from a growth standpoint, but the customers that are moving off their sand aren't necessarily moving to another tech protocol. What they're doing is moving to a service provider and the service providers for the most part are running on ethernet. So yeah, the functionality that's inherent with ethernet is much more scalable and has many more robust features when it comes to the agility. Literally our sands virtualize and we can literally shuffle around at an IP level the virtual machines not only where they're at but also their target on their sand. In a fiber channel environment that's very constrained and the technology set just doesn't seem to be coming there as fast. So if you're looking at a roadmap of adoption, now that the VDX has had the latency that everyone's bragging about, the .3 mils, now that they actually have the speed 10 gigs, 40 gigs, the collapse is going to happen and now as a service provider and you can make those moves ahead of time, commoditize that and actually monetize it as well, it makes sense day one to do that. Okay, so your customers are service providers. One of the biggest questions we have about cloud adoption in general is, where do the applications live and what does a hybrid cloud look like? So the question I have for you is, are there some applications that customers are keeping in house, making private cloud environments? There's other applications that they're saying, I'm going to go to some as a service type solution and the question is, are there applications, are there's two separate ones that I just need to maintain or are you actually seeing things like cloud bursting becoming reality today? You know, what we find is the main limitation on all this is just how fast literally you can move one packet from one location to another and the speed of light's not changing. And it's actually a lot of our growth strategy is putting these micro-clouds in all the major cities because we find out that these enterprises are on three and five year adoption cycles, but their applications and their cluster demands aren't following those adoption cycles. If we can actually give them an agile infrastructure to do cloud bursting, they'll do it if that latency, if their core applications will respond out of house as if they were the same in-house. The easy ones of course are DR, course are development. But what we find out is where you actually derive value and where the organizations are really adopting is actually in production. As long as you can achieve that sort of latency limit there. Okay, Dave Stevens in his presentation this morning, Brocade CTO said that distance doesn't matter anymore. And Brocade has some nice WAN solutions, but what I've heard from you and I've definitely heard from others, latency is so critically important. Flash type solutions on the storage side, if I need mobile applications and VDI remotely, latency is critical. So how are you helping to overcome the challenges of the speed of light? Sure, so I think you did have a correct point. I mean, our site to site nodes are 40, 80, 100 gigs of literal throughput that we can put between our data centers. So in that aspect, it looks and feels as if it's local. But the reality is, is if it crosses a couple states and it's 10 milliseconds, that itself is a time-induced latency on the applications. And our growth strategy, as I pointed out early, is to drop these cloud nodes in the local cities that sort of have the geographical regency for there. So if you're not just in Seattle, you're in Denver, Chicago, New Jersey, Texas, say Austin and maybe Dallas, because that state's so big, where you can actually then cloud burst into that node that's close enough so the latency is maybe in the three to five millisecond range. And that's uncomprehensible. That's a couple of bats of your eye. You're not going to notice that. So from a speed of light perspective, we actually think that's the delineator and the determinator that differentiates us on the market. And Brocade is doing some great things about, and just got done sitting through another session on the next line of VDX on driving the latency down so that now the virtual routing and all the aspects of routing, all the aspects of virtual MAC addresses will literally just propagate out to the first top possible. So now you can actually make one decision point but push it out throughout your entire nodes. But still as an enterprise, when it comes to adoption, they have to get just that speed of light variable as low as possible. Okay, so Luke, it sounds like you use technology as a differentiator in your business. You adopted kind of the trill VCS functionality early. What technology areas are you looking to for kind of your next spin? We're almost looking at trill as a base fabric. And let me say, you don't have to limit yourself just to the network side of things, that you've done some cool things on the storage side too. And I think that's the interesting piece. What we've tried to build is just an absolute fabric. And I mean a storage fabric, a network fabric and then a compute fabric. And now what I'm really excited about is what fabrics are you going to build on top of the fabrics. VXLAN is just another fabric I've built correctly on top of the fabric. The NetApp cluster mode, it's now a SAN, but now we can actually literally abstract that and map that across any of these fabrics. And now we can integrate it into the next level of fabric. If the SAN can now look at VXLANs, customers will literally be able to drag and drop in a virtual data center. What connection from the host all the way to their virtual connection on the SAN. And that's a delineator from our standpoint and how we deliver our service. So our providers and SaaS providers can roll out their applications faster to the market. Okay, and so how do you manage and orchestrate all of these various technologies from the pieces of the stack that you discussed? That is definitely a key piece and something that we're thriving and trying to look for on top of it. That there's still not sort of that wrapper that comes across any of them. And there's not some open API stack that comes together. And really what you end up doing is picking a best of breed, whether it's Windows or it's Hyper-V, VMware to manage up to a particular point and then another provider sort of matching that. We are looking at some very new monitoring level applications. I can at least give us a homogeneous look from all the different sides so we can understand failure zones, hot pockets, that whole scenario. But from a pure orchestration layer, nobody's really mapped it into end quite yet. Okay, and do you have some opinion on OpenStack? I definitely have a great opinion on OpenStack. I think the development on it is going great. We're actually currently even looking at ways to adopt it. Our current customer base has a particular level of shyness to it. And definitely, I think the leaders in the OpenStack community also might sort of be hindering it because they themselves are inhibiting other people from entering into it because they're seeing such a forerunner into it. So I think as that community actually builds out a little bit more, some other thought leaders sort of come to the head on that. I think it will allow a much rapid more adoption from the enterprise side into it. Okay, so last question I have for you is that companies that are adopting the solutions that the service providers that you work with are providing, what's the profile of them? How are they, is it a leap? Is it, do they have to get the CEO buy in? What's the profile of that person in? What's kind of some of their biggest challenges? That they're facing. Sure, so interesting enough, our end user adoption seems to be coming from the C levels. So it's a CTO or CIO understanding the transformation that's happening. And then just as important, it's the CFO that's understanding the transformation that's happening at his level. I mean, he can move those acquisition cycles, the operational cycles into that monolithic three or five year block. And now he can actually start as a CFO, work with the CIO and derive value from an enterprise saying this application is going to go through six month iteration cycles. So why doesn't our infrastructure, why doesn't our IT have that same level of agility? If that decision's made at that level, it's a rapid, rapid move to us. From an IT level, it's usually a toe dip still. We get a DR, we get a development. And our favorite line is your development is going to run better in your production probably in-house. So we'll take it, because six to eight months later, you'll start feeling much more comfortable migrating that in. Wow, is it like rather than the consumerization of IT, it's like the dev-op, you know, taking over that. Exactly, and they sort of drag everything with it. Okay, so you've been talking to a lot of companies, you know, step outside of your current role. You know, what excites you about technology today? What have you seen that just kind of, you know, says wow to you? You know, what I'm seeing actually is, and where I'm just saying wow, the incubators, the startups are being able to bring products of market faster because of cloud. But also what you're seeing happen with this whole software-defined methodology is a company like Brocade does an amazing job defining ASICs and hardware. And that is a life cycle. They actually have to develop it and print those and push those markets. As you take solid foundations, but yet you move it up to software-defined levels so that these rapid adopters, these cloud companies, the startups can now write code much, much more faster. You're going to be deriving much, much faster levels of innovation all the way through the stack. So I think in the next two to three years, once again, as that foundation gets more solid and the actual feature sets get pushed to the edge, the software levels, you're going to find a much rapid adoption. You're going to find new technologies, people wouldn't have thought about. You're going to find a much easier level for end users of people to sort of manage their solutions, but there's still going to be that core that's just going to be incredibly, incredibly complex. And I think we've heard that today. Well, it's interesting. So my closing thought on it is, say, right, it's complicated. And Brocade's trying to simplify things. I was actually talking to Lisa Kaywood here at Brocade and saying, you know, Ethernet fabrics and SDN, you know, how do we mesh those together? And Ethernet fabrics gives that simple foundation. I've got solid hardware. It works as you keep saying, you know, I grow, I add a switch, I add a new switch. I upgrade the software and it just works and upgrades so much simpler than what you were doing before. And SDN is going to allow me to control that. So, you know, simplicity in, you know, a solid foundation. And then I can really customize through the software feature. So, you know, Luke, appreciate you coming on theCUBE. You were always welcome at our event. So we've done, you know, calls with you, CUBE with you. And, you know, thank you for sharing with the community all that you're doing. Thank you for having me. All right, so this is Stu Miniman with wikibond.org and we'll be right back with our next guest after this quick break.