 Live from Vancouver, Canada, it's the Cube at OpenStack Summit Vancouver 2015. Brought to you by headline sponsors EMC and Joyfully by Red Hat and Cisco with additional sponsorship by Brocade and HP. I'm Stu Miniman with wikibond.com. Here were SiliconANGLE TVs live coverage from OpenStack 2015 and beautiful Vancouver right here on the harbor. Going into this segment, talking about some of the networking pieces with our sponsor, Brocade. I've got Dave Meyer, welcome back to the Cube, Dave, Chief Scientist from Brocade. Thanks, Stu, it's good to be here. Great, and Tom Nadeau, it's your first time on the program, thanks for joining us. Distinguished Engineer with Brocade. Thank you, great to be here. All right, so gentlemen, last year, one of my disappointments in Atlanta was just when I got all the feedback from the community and dug into it, it was neutron, was not nearly as stable as most people would like and a lot of things going on. A lot has changed in last year, so before we dig into that, maybe introduce to our audience what your role is at Brocade and what you're working on, Dave, we'll start with you. Okay, so what I'm working on right now is, well, I had been chair of the Open Daylight technical steering committee last year when I was on the Cube, we spent time talking about that. Right now I'm the board, I'm the chair of the board. Our colleague Colin Dixon is chair of the TSC. Kind of maybe unusually right now, I'm working mostly on something called machine learning and applying that to networking, which actually fits well with all of this, but it's sort of a little bit of a different area. All right, so Dave, yeah, we did an event with the MIT Sloan School looking at the second machine age and the whole digital economy. Let's not talk about OpenStack, let's talk about that. All right, Tom, let's talk about what your role is at Brocade and what brings you to the OpenStack show. Yeah, my role at Brocade is, I've got a couple roles. I'm sort of half in the CTO group, a little with Dave and those guys, but my primary role is to run the development teams for OpenStack and Open Daylight and OPNFV, although they're all kind of related. All right, so if the premise coming into this segment was last year, Neutron needed a bunch of work, I know much has gone on, you guys are helping, the community is helping, a lot of people involved. What's the state of networking in OpenStack? You want to take a first shot at that? Yeah, sure. So I think things have evolved a bit, but I think there's still work that needs to be done. I mean, we've done a lot of work in ODL where we've gone in and sort of brought things up to snuff a bit. The project got sort of left alone a little bit when one of the vendors kind of departed a little bit. So we've come back and rebooted that project a lot, spent a lot of effort on that, but it's kind of an iterative thing. So we've spent time sort of getting things back, getting people together, getting people working on it regularly. There's two guys on my team that are working on this regularly now and there's guys from Red Hat, guys from Intel and a couple of other folks I think coming in. So I think that's good, but I think there's more work that needs to be done between this sort of Neutron ML2 sort of interface to ODL versus what's going on on the southbound side, getting all that sorted out, plus the group-based policy or policy situation. How does that go in here? There's, you know, so it's going to take time. Yeah. So Kesha, one thing about that is there's kind of a meta angle on this too right now is that one of the problems we had in Open Daylight and one of the reasons why there was sort of a divergence between Open Daylight and basically Neutron and this is the thing you've been talking about that's been kind of, it's getting better is that from a perspective of the Open Daylight community we weren't really participating with the Neutron community, right? And so that has changed and then the Neutron PTL, this guy named Kyle Mischer is a really great guys, really been fairly active in Open Daylight or at least helping the Open Daylight Neutron crew, you know, kind of. Oh, Kyle's on the team now. Yeah, yeah. So you know, there's this good integration going on which we didn't have last year. Yeah, no, great. I mean, yeah, for those that aren't familiar, you know, originally it was really the Neutron guys that contributed a lot of the Neutron code. Talked with VMware, we're going to have, you know, some VMware people, you know, definitely on the queue plenty of times. And you know, they are kind of put a renewed effort over the last year into various pieces of OpenStack, but from Neutron standpoint, it's a whole lot of those people. I know Kyle from his Cisco days, he's now over at HP with the Helion group. What is that relationship now between ODL and you know, what's going on in Neutron? How much, you know, sharing of code is there and how much is it just kind of relationships between those two open source bodies? I'll take it for, I'll take a whack at that. So like my perspective on this is that there are at least two camps, you know, there are the people who think, well, Neutron should be able to do all of networking and we were just having an in-depth discussion about this last night over dinner. And then there are people who think, well, there should be some piece of software or some agent mediating between Neutron and the network. And what it came down to in our discussions last night and previously is that there's sort of overlay and underlay management that's going on and what Neutron is good at is overlay management, right? And so the ideal role for something like Open Daylight is sitting between the two. So it has visibility into the overlay and the underlay and it has a kind of a unique position there. Yeah, exactly. Yeah, I think that was the conversation we were having, somebody asked this, so what role does ODL play in an all open stack, say VXLand deployment? I said it has lots of role to play. I mean, it's got to manage the underlay. So in a lot of ways it handles the traditional NMS, EMS functionality that you used to have with a bespoke kind of application. But then like Dave was just saying, it also sort of joins the two together and lets you bind them together if you want. You don't have to either, by the way. Yeah, I mean, if I look at it, ODL has always been pretty good at kind of the southbound. That interface down with the hardware, really saying what's the controller going to be and okay, how does that work with a variety of platforms? So that's something I think if we can take that, learning from ODL and bring that over to OpenStack, that's goodness. Yeah, very much so. And just to not leave you with the impression that everything is perfect there, there's a bit of a philosophical divide with respect to this because there are, I don't know if you want to call them purists or people on one end of the overlay spectrum believe that the overlay should know nothing about the underlay and that the value of the overlay is in just that decoupling. There are other people who go, well, you need to understand what the overlay is, otherwise you can't debug it, you can't optimize it and all of this stuff. And so there's controversy there yet. That hasn't shaken out completely. So ODL is sitting in the we do both camp and there's controversy around that. So it's not, I mean, there's evolution that we're going to see over the next year in all of this. I mean, the way I was describing it the other day was this kind of this continuum, right? There's sort of all the way over here is the sort of the applications. You don't need to know anything about the underlay. The underlay you plug in and they're just stupid ethernet switches, right? And then there's all the way on the other side which is you need to tightly bind the two together so that one can control the other and all that. Wait a minute, there's not ATM or anything? Yeah, I know. And so there's this continuum, right? And I think the cool part about the ODL OpenStack integration story is you can actually slide that around depending on how you want to do it. It's not prescriptive in one or the other or whatever. I mean, you can slide this around and meet your needs. And we have customers that are deploying ODL with and without OpenStack and it's a different sort of facet on that spectrum. All right, so if I could step back for a second here. From the networking space in general, one of the big conversations we've been having is that whole kind of software-defined networking. How do we simplify networking? How do we make it much more extensible? The open networking user group, the user community actually said, I'd love to have a software ecosystem that I can do more like what I do with the rest of kind of my purchases. Is there an intersection between kind of what's going on in SDN and here at OpenStack? And if so, how do those mesh? You want to go? Yeah, so I definitely think there is one. I mean, we are actively building one for our applications that we're building on the controller. We're very much treating the controller like an operating system. Very much, there's kind of infrastructure, fundamental components that go in there and we're not going to reinvent them in every application that we build or other sort of pseudo-infrastructure components. And the cool thing is you can mix and match these applications as you need to build them. You can choose to build your own, you can go buy them. So I think there is the same opportunity happening that happens in OpenStack in the SDN controller space. Yeah, I agree with that and the only thing I'll add to that is that I'm a little bit, I guess it's ambivalent is the right term about the prospects for the ecosystem looking sort of like the iPhone ecosystem with the app store and all that. It's not an app store, yeah. You know, it's, that to me, as I say, I'm ambivalent about that because the network space for, at least for infrastructure components isn't like the app space for your iPhone, right? So maybe there'll be applications, but there will be applications, but it may look more like a Red Hat sort of model or something like that where basically you bundle value-added applications with whatever the infrastructural components are. Well, it's a little different for one important reason. You know, in the app store there's sort of small, generally speaking, the kind of small apps that are not tested to interact with each other a lot. This sort of interacting with the base system. There's maybe some security checks and all that and the light testing. The applications, I mean, you got to remember the controller and open stack are infrastructure for network and services deployment. And so there's a significant testing component that needs to go to pre-qualify that, like the Red Hat model that you're saying, where if you're an officially qualified Red Hat app or even canonical, those guys have a whole testing infrastructure. You got to plug into first, get qualified, and then you can be offered. So it's a little different. Yeah, yeah, and the point I guess if I look back at the keynote this morning, Jonathan talked about if we have interoperability and potentially even federation between certain environments in an open stack environment, I can buy software and not know, need to worry about kind of the underneath is the software I buy going to work on this component. Actually, it brings me to my next question is, we haven't talked about containers yet. One of the real hot topics we've really had in the industry the last year or so, and now there's Magnum and a couple other projects here in open stack, one of the big gaps in the whole container space has been networking. Of course, Docker bought Socket Plane. Give us your thoughts. How does containers, open stack, networking in general, what's your position on that space? All right, go ahead. Yeah, well let me give you my two cents and then Dave can give you his. I mean, the space is ripe with potential, right? I mean, you can containerize anything from the controller to the open stack components and for a lot of good reasons. But the power in the containerization in a lot of ways is for applications, right? In that their packaging is sort of irrelevant to what they run on. That's one of the big advantages of container, right? Where you can package up the app and it'll run whatever, right? And that gets back to the previous question, right? Were you testing and qualifying these things? So I think I almost don't, in that space, I almost don't think containers change the equation, right? You still have to test and qualify those things. It's just a different delivery mechanism, which is potentially faster, cheaper, but still containers also have their own security issues, right? So you can't just flick these things around willy-nilly, right? You're going to have to still do the hard work. Yeah, unfortunately we still haven't found that silver bullet for any of the IT problems, right? Yeah, yeah. Well, lucky thing for us, right? Lucky thing, right? I'll just say one other thing is that you mentioned the socket plane guys, well, Brent and Madhu and Dave Tucker and those guys, they did see, I mean, the kind of brilliant thing they did was see that this technology was coming up and that it's networking capabilities were in question, right? And a lot of the work they've done, I think it's improved that. But like Tom said, if you kind of think about it this way, how long have we had, I mean, how long have we had hypervisors around so that we could have VMs? It's been a while. So it took a while to get to where we're at and even there are issues that you could raise. So the technology is fairly new. I think one of the things that really attracts people about is that it's just much lighter weight. Yeah, operationally it's easier to deploy and kind of manage, but. Well, every aspect of it, basically, right? So, but we have to solve many of the same problems. Yep, all right. So Dave, what are you hoping to get out of the show this week, what's Brocade working on and what's exciting you in this space? Yeah, so a great question. So, I don't know if we're talking about this on or off, but I've been working on sort of up-stack things these days, machine learning in particular, and what's really interesting, I'm going to say this here, maybe I don't want to do this, but. You know, what's really interesting is that this technology is really, really powerful and the network space where we work, where we spent all our life working are the last couple of decades anyway. I hasn't been touched by this. So, if you think about what is orchestration, what is, even what is control and management, you see that there's all of this opportunity to work in the, to put these two spaces together. So one of the things that I've been doing since I've been here the last 24 hours or whatever is, I've been talking to people about what their use cases are and what their pain points are around open stack. And what I find is, you know, that if you think about an automation continuum of which open stack is a big part, there's a really nice opportunity to do some interesting work here on top of open stack or on top of ODL. So that's kind of what I've been doing. Yeah, I was going to say ML actually has some interesting potential. Like we were just talking about containers, right? And the sort of the logic or intelligence behind deploying the containers, how, when, and not just how and when, but adapting as you go, like detecting security anomalies as you go, detecting these patterns, right? Those are things that don't exist in these systems today. I mean, if you look at the way I look at the open stack, ODL, OPNFV, all these things that we're building, these are still very primordial in the sense that they still, yeah, they're automation and that's one of the cool things that SDN has brought to the party is the automated programmability. But you can make that faster and you can also make that better, you know, with ML if we can figure out how to wrap that sort of, you know, we keep using the word intelligence, but you know what I mean? Around that. So Tom and I have been working on some applications that use the massive amount of flow data that's around. I mean, there's just so much flow data everywhere and we have some kind of interesting applications. You'll see these after a couple months or so. Well, you know, we definitely look forward to catching up on some of that, the machine learning stuff, you know, going forward, love when we can take the, you know, infrastructure and help move it up stack to really drive some new business value there. We're going to have lots more digging into the networking space this week. So, you know, Tom and Dave, thanks so much for joining us. We'll be right back with lots more wall-to-wall coverage here from Beautiful's Vancouver at the OpenStack Summit 2015 after this quick break. Cool.