 Hi, this is Stu Miniman with wikibond.org. Here with Silicon Angle TV's live special coverage of Brocade Tech and Analyst Day here at San Jose, California. And we've got a couple of kind of deep thinkers in the networking space where we're going to go talk into some of the reality. So sitting right next to me is Greg Farrow. So people online probably know him for ethereal mind is his blog and he also runs the packet pushers podcast where all the true networking geeks go and argue protocols and what's happening in the real world of the networking space. And also is John Hudson, who's the solutioner on Twitter, I believe. And he's also a global architect for Brocade. Gentlemen, thanks for joining me here on theCUBE. Look, he's real. He is real. So, you know, I've had the pleasure of knowing Greg for a couple of years. So, you know, Greg is one of those people that online sometimes he tells the truth and that sometimes annoys people a little bit. But in reality, he tells the truth. So it's one of those people that when you get to know him, he's still Greg, but you know, everything there. It makes him very useful. So, you know, there's so many topics. It's really almost like a renaissance in the networking world. Oh, he said, you know, it's fun to be networking again. It's absolutely fun to be networking. Jason Roulal, you know, great networking sexy now. So, you know, it's exciting. But, you know, Greg, one of the things I've been arguing with you about a nice friendly debate is, you know, networking is not the center of the universe. When we go to the software-defined data center, you know, there's big pieces. You know, come on. It's the plumbing. What's the role of networking today? And how is it changing? Okay. So in my view, networking is the definitive platform. So forget VMware or Hyper-V or Citrix Zen or KVM. They're just software platforms for operating systems underneath all of those. Yeah. They're just package generators. Right? It doesn't really matter what the package generator. The network is the platform that holds all those things together. And it's, people tend to forget that because for the last 20 years, we've had these really crisply defined, well-controlled edges between the upper layers and the lower layers. And then in the last five years, we've seen that edge collapse. And so as VMware came along and the hypervisor and created a dynamic capability, that broke down those crisply defined separation between the network and the applications that ran above it. So now the network doesn't just see applications as where the service plug-in to the ether net ports. We now have virtualization and even deeper right the way into the application stack. So the whole purpose of networking today is that we're now the platform that supports a dynamic system or dynamic ecosystem instead of the static system. And that's what's given rise to this SDN topic. But it was what we always wanted. I remember thinking how ridiculous it was that I'd run an application, the application would complain. I would take that information, go to the network, teach the network something. The network would do something to the application, the application would complain. And it's like, why am I in this equation? Right? Why isn't the application just telling the network what it wants and the network telling the application what it can provide? Yeah. And in the past, we've always been able to rely on autonomous protocols. So we've had protocols like OSPF, ISIS, BGP, MPLS, and LDP. And those protocols all worked in a defined way, but they happened all on their own. So the network went in just some little magic around the CLI. The network self-configured itself according to the rules of the protocol. And then, bang, we were done. And then if the network broke, then it self-healed. If the network needed updating, it self-updated. If the bandwidth changed or whatever it was. Now we've got a situation where those autonomous protocols are being impacted by the layer above. And that's where the SDN comes in. So an open flow, I think SDN's been coming for a long time, really, hasn't it? Well, I mean, you could argue that a lot of what you see in some of the larger networks, you know, the AT&T's the world and so forth, some of the tools that they've developed internally for how they manage very large situations are really precursors to a lot of what's being done now. But they were just, they were done in very proprietary internal specific ways to them, right? And so now it's, how do we bring that out? Because I mean, I'm not only about you, but when I first learned how routing worked, I was very disappointed. Like, I thought it was much more advanced than that, you know. And so for me, SDN is almost a fulfillment of what I always thought routing was, right? This idea of this very kind of morphable, aware layer that can react to an environment, right? Yeah. Now, getting from A to B isn't always as easy as it may appear on the cover of a magazine. But this whole concept of how do you teach a network new tricks, right? So Greg, you were poking a little bit at VMware. What's your take on the kind of virtualized pool and automate when it comes to networking? Is that the right methodology or is there a different way we should be looking at how to change networking? Yeah. So you've got to look at networking as networking, not networking as virtualization. So most people are framing the networking debate in terms of what, so VMware's doing a much better job of whoring itself into the industry and saying, we're the only solution and you should only operate inside of our platform. And VMware tries to market itself as a platform on which everybody else speaks. Yeah. You know, I would actually say that, you know, yes, of course VMware like every other, you know, publicly traded company wants to rule the world, but they're actually embracing some of the multi-hypervisor and even physical worlds. If you look at the dynamic ops acquisition, then they see your acquisition, but you know, they're trying to kind of put their arms around and pull it in. Yeah, but to your point, VMware is now a very, I mean, I remember friends complaining when they were at 6,000 people, right? Now they're like 13,000 plus, right? So you now have groups doing, I mean, I did a presentation a couple of weeks ago with a great guy, Jad, from the Navy team for VMware and then who was the, at the time, the VP of Federal Sales for NYSERA and it was right before the acquisition became official, right? And so the guy from NYSERA is up there talking multi-hypervisor and, you know, lots of open environments, right? And then the guy that was from original VMware was able to say, yes, but was in that construct, we think that our version of it is best, right? So you can have some coexistence of this idea of we'll allow you to pick whatever flavor you want, but chocolate's best and this is why we think chocolate's best, right? Yeah, I think VMware sort of realizes that IT infrastructure is actually bigger than VMware. VMware's got a nice slice of it, but they can't beat Linux. Microsoft couldn't beat Linux. They couldn't defeat Linux and the open source movement. Well, it's interesting to say that. So, you know, I talked to Martin Casado, one of the co-founders of NYSERA and I know you saw the interview we did and he actually was banging on the table, you know, we're going to change networking and he said there's open source projects and there's open standards and there's room for both of them, but I hope the standards die is basically what he said. You know, what's your take on that? That's nice. I'm sure he'll enjoy that. The open source usually self-explodes. It just reaches a certain point. It requires a benevolent dictator to hold a coherent story together. So, in the same way that VMware has managed to make an ecosystem by having a consistent story, develop a product and they've managed to, you know, own a certain percentage of the industry, the only reason that Linux has been successful in all of its various forms in the open source movement is that somebody becomes the benevolent dictator to hold the pitch. Okay, well, it gives you an 800 number that you can call when things go wrong. Yeah, so, the line is 12 volts, doesn't write the whole kernel, but he holds the kernel together. All right. So, do you not buy that then VMware can be, you know, driving an open source project and, you know, has the skills for that now? No, EMC doesn't have that sort of DNA. Well, I said VMware. Well, EMC is VMware. Yeah, and I would agree that. And VMware is 100 in EMC company. So, EMC wants to promote the picture that VMware is. At some level, a company may try very, very hard and very honestly to support an open source movement, but at a very fundamental level, that is a cost center. That is a business unit expending company resources that generates no perceivable income. It may generate goodwill. It may generate indirect income, but no perceivable. And so, it's very hard for them to survive through the ebbs and the flow of a business cycle. When things get tough, open source movements get hurt. John, so you work on standards and, you know, so isn't it the same? I mean, standards is one of those things, you know, it's kind of boring to most people. They don't understand it. It takes years to do it. It's a time sink for, you know, the vendors involved. Yeah, but you've got to realize that there's really three different groups of us in, for example, the IETF where I work, where you have the academics, right? You have the people that are there for the love of the algorithms and the love of the mathematics, right? And then you have the real vendor folks that are really there to use standards to kind of take advantage of vendors. And then you have another group that I consider myself part of, which is people who used to be operators who know what it's like to be abused by a proprietary protocol and really fundamentally believe in standards, right? Like, don't see it as a positioning thing. This is how you protect little people, right? This is how you keep people from being abused. A network is far bigger than just VMware. VMware is a minuscule little animal that sits in this little data center space, but networks are global entities, Internet. They're... They're interlocked with each other greatly. And they're all interlocked in a way that VMware can't, that a hypervisor system can't. Like, vCenter doesn't scale beyond 50,000 nodes, whereas the Internet is already a half a million nodes. And I'm just counting the routers, not counting the, you know, the edgepoints and all the computers that connect to it and that sort of stuff. You can't have open source in that because there's no one person who can carry the initiative in a sustainable way. You have to have a team so that the banner can pass from person to person to person, and that's where open standards supersede open source. Now, if you've got a tightly bound problem space like Python or PHP or the Linux kernel, which is nice, then you can have a dictator, you know, a dictatorial type of capability, which can focus and take leadership. Often benevolent, but a dictator on us. Yeah, or a fundamental core that can hold the vision together over, as you say, over the efforts and flows of the business cycle. One thing that we know about networking and open standards is, 30 years later, they really, really work, right? So how many open source projects have lasted 30 years that have been taken over by leader after leader after leader. And that doesn't always mean you get the best solution, right? Sometimes it means that it's better to pick the solution that isn't actually the best solution if it means wider adoption, right? I mean, you can look at a lot of different protocols, you know, 802.11 and others were essentially, you know, I'm not saying that there weren't better solutions to the technical problem presented, but for whatever reason, that instance of it became what enough people wanted to use, that we benefit more from the wide adoption of it than we do from the perfection of it as an individual. Okay, so guys, with the limited amount of time we have left, I want to bring us back to today. Both of you talk to, you know, Greg, you architect a real solution, and in your community, you know, are the folks in the field that are putting there, and John, you know, I know you're talking to those end user customers. So, you know, want to run through a couple of topics, and you know, where are we today? So Greg, I'll start with you, you know, the whole, you know, layer two multi-pathing, whether it be Trill or SPB or other solutions, you know, what's the maturity of that today and what are you hearing from people deploying it in the field? It works, people are using it. And I think we're going to see a lot more of it. I think we sort of got through, you've got to remember that Trill's already 10 years old, just close to it. Yeah, yeah, yeah, yeah, yeah. Nearly 10 years. Yeah, yeah, getting there, yeah. The Trill idea has been wandering around for a decade. We're finally getting to the point where there's actually a clear, so somebody's been quite prescient and rowdy of people and led the first years, you know, the initiative. Five years later, we're finally, you know, 10 years later, we're finally into the point where it's met the requirement. So a lot of these people can envision these problems long before most software companies have even envisioned that there's actually a need. And you have a nice common thread between the three of them, you know, for me being SPB Trill and some kind of an embolag type scenario, right? And that you're trying to kill active passive links, right? I mean, if you paid for it, you want to use it, right? So the Layer 2 multipath fundamentally changes the equation. Half of our data networks today are dead because they're redundant paths for something else, but they shut down and unused. So 50% of your networking asset is dead money. So Layer 2 multipath lights that up. Soon as you mention that, everybody goes right, how do we get onto that? So people are starting to deploy that. Cisco's been pushing their fabric path along, but Brocade's VCS is quite interesting to me because the automation that they put into that is quite, there's a lot of uniqueness into VCS in terms of what it can do in terms of automatic load balancing, back pressure, link bundling, and so forth. And the funny part about it is if you actually configure a VCS, you kind of just plug it in and say VCS there and it's done and there's all this magic happening under there and you don't know about it. But if you're a networking nerd, you just sort of scratch away it and you go, wow. And then you scratch away it and go, oh, they thought about that too. Hang on, and that, why don't you? So John, the customers you're going to see, a lot of times customers are reticent to adopt new technology. Is there anything that you said, I just wish if they would all deploy this feature or activate this, it's going to make their lives a lot easier. What aren't customers doing that they should be doing today to make their lives easier? I would say that what they're doing a really bad job of in general is handling the human process element of it. The idea, the concept to touch on VMware, of telling a VM an infrastructure, you need these resources and if you can't get them here, you go elsewhere and find them, it's fantastic. But the truth of the matter is, if most IT directors walk into the data center today and ask where's the Oracle server? And the admin went, ah, I don't know, last week it was over there. It wouldn't go over very well, right? So there's a mind gap between how much nimbleness is available and then how comfortable people really are in a world where you're used to having a data center number with a rack number, with a shelf number, with a MAC address, with an ID, with a software license. And so there's a lot of great technology available and people are having a hard time adopting it because of change control policies and all the different things that we've kind of wrapped around IT data centers and so it's making change very difficult. The other problem here is that management needs to change. So the actual IT management people need to come and meet the technology people as much as the technology people need to come and meet the managers. Right now we've got a big problem in networking where network managers are saying, I want this, but I'm not going to do anything to get there. I've either got to spend, I've got to change my thinking, do something different. So Greg, I've got a bunch of service providers here and Brocade seems to be making some good end roads there. I believe you have some maybe concerns about, just moving our network kind of out to service providers. Do you want any commentary on that? I'm not 100% sure what you're driving into there. So I'm saying, should I manage it in-house and change the way I'm doing things, or do I just go to a service? So right now I believe that the only way you can make data center networking work for your company today is to do it yourself. Because there's so much going on in the data center space, there's so much vectoring of change, the speed of change is so fast and the market's changing attack. If you outsource that, you lock yourself into a technology which might be a dead end and you could be. But that requires that IT organization to recognize that they can be a benefit and not just a treading water model, right? Looking for ways that an IT, you look at someone like a Netflix or someone where their IT processes have really made them more competitive in their market space. And if IT organizations aren't stepping away from just keeping the lights on and really looking for ways to innovate and become better assets to the organization, then that outsourcing is a lot more likely. I guess my point is, today there's so much change going on and it can make a big difference to your business. You need to do that, you need to be very close to where you are. Now for example, 10 years ago, say 2000, 2001, you needed to do your own WAN because you needed to customize that WAN to meet your requirements. But somewhere in the mid 2000s, WANs became commodity standardized flat. So I'm perfectly happy to take my WAN and give that to an outsourcing provider because there's no innovation happening there. I can't, it can't deliver any value to my business. So just give that away, maximize the price, minimize the cost, maximize the benefit to the business. It's static, it's unchanging, outsource that, right? Desktops, outsource that if you're going to stick, unless you're going to go into VDI, whole new argument, right? In data center networks, you've got to realize that there's a whole great opportunity for change going on. So outsourcing your data center. You're going to miss it. You're going to miss everything. So even if you push your stuff into Amazon, there's so much speed of change happening in the Amazon space. You're going to kill yourself because you're going to take a stake in a technology set and then you're going to find that Amazon is different. And then you say, you're going to say, you're going to say, you're relying on them to be innovative, right? So guys, we're running out of time. There's one thing, you know, I mean, obviously the big news of the week was the iPhone 5. Second, you have this, the VDX8770. I can't order it. I tried to order it. I tried to order it till Friday. Two more days, you'll be able to order it. But IDF's going on this week and Intel made an interesting network announcement. Talk about really taking the fabric and pulling it in. I mean, Intel's, you know, for years been working with a broad ecosystem and slowly moving functionality in. They made some big acquisitions with Fulcrum Micro, Infiniband from Juniper, you know, the company that... Q-Logic. Yeah, Q-Logic, sorry. Thank you. And so, you know, Greg, first take on, you know, what's Intel doing and what's it mean for the networking world? Meh. Meh. Meh. So what, you know, they put DPDK into the chipset. All right. I would argue that it might be in all of our best interests to go to Intel, ask for forgiveness and hope that they're benevolent when they finally take over your market. You know, they are a phenomenal organization with a lot of really brilliant people. And if they want to get good at something, they will. And so I think it's very interesting, but I also think it's a little scary. Yeah. You know, it's... It's interesting. Of course, you know, VMware's moving to take over more of networking Intel now, moving into that space. There are lots of companies, you know, it's not just Cisco and the traditional networking companies. It's coming from all sides of the town. But if anyone's going to build SkyNet. Yeah. If you want to prognosticate, so imagine if Intel uses the DPDK extensions to actually build a switch inside the CPU. So how are you going to configure that switch? Are you going to do it with VEB VEPA? Or are you going to use something called VEX model? Yeah, yeah. Or are you going to use some other, you know, VN tag? Which is where some of the companies go wrong because they get started from technology. They don't pay attention to the OAM, the care and feeding. How are you going to administer it? Are you really paying attention to the workflow models of the customers you're trying to approach technology with? Yeah. So if you imagine today, Cisco's got their MKA chipset going on, the Palo chipset on their adapters in their fancy UCSs. So if you imagine that chipset coming into the Intel CPU, what does that do to the server market? So instead of actually having a fancy adapter, there's actually a switch inside the Intel chipset. But how are you going to configure that? How does that integrate? So my question is, meh, because Intel's not laying out a strategy to say, this is how we'll integrate with VMware and Citrix. And what does that do to competition and innovation and other things, right? I mean, you know. I don't think it changes. It changes what happens in the hypervisor in the sense that instead of having a virtual switch, or instead of using a network adapter to provide virtual switching functionalities, there'll be some subset. And I imagine that in long-term, it'll be like the Intel graphics adapter. Yeah, yeah, yeah. You get the... The bare minimum, right? The low performances on the chip is in the CPU. And then you put in a fancy one outside. You need the high-end, you augment. So yeah. I imagine that in the end... And it's not a bad model. You know, perfect. And you might be able to say, if I'm Google, I can make the most of that. I can make that fit my business model, do some fancy programming in a way I go. And but if I want some, I want to be in the enterprise and I want it to stand up and beg like a little doggy and then brush my shoes and give me a hairbrush in the morning, then I'll go and buy a fancy adapter and the market will be unchanged. That'll be exactly the same as what we've got today. So guys, we are running out of time here. As we wrap up, what I really like is, I've had a lot of these conversations with you guys, online, through blogs, through TwitterSphere, in person, and we're going to keep the conversation going online. So Greg, you're like the Howard Stern of the networking, you're the king of all media. You're found on network computing, ethereal mind, a packet pusher's podcast, and he does drop an F bomb every now and then. So where's the best place for people to find you? You can find me on my blog at etherealmind.com or on the packetpushers at packetpushers.net. I'm on Twitter as at etherealmind. All right, and John, you're blogging on the Brocade site. I occasionally get something published on the Brocade community site and as you mentioned before, the solution here on Twitter. And if you buy enough, I get to come visit you. Excellent. So gentlemen, thank you so much for joining me. This is Stu Miniman with wikibond.org, here with continuous coverage from Brocade Tech Day in San Jose, California. We'll be right back with our next guest.