 San Francisco, extracting the signal from the noise. It's theCUBE, covering VMworld 2015. Brought to you by VMworld and its ecosystem sponsors. Now your host, Stu Miniman and Brian Gracely. Welcome back, this is theCUBE, SiliconANGLE TVs live production of VMworld 2015 here in Moscone North, San Francisco. Happy to have back on this segment. We're actually going to dig into some of the networking pieces of Brian Gracely and myself here hosting it. Sean Walsh, repeat CUBE guest in a new role though. So Sean, welcome back here now, the general manager of the Ethernet business at QLogic. Thanks for joining us. Thank you, thanks for having me. All right, so Sean, we're joking before we start here. I mean, you and I go back about 15 years. I mean, those that know the adapter business, I mean, J&I, the J-Core business, you've worked for QLogic before, you did a stint at Emulex, and you're now back to QLogic. So why don't we start off with that? And what brought you back to QLogic? What do you see as the opportunity there? Sure, I'll tell you, more than anything else, what brought me back was this 25 gig transition. It's very rare, and I call it the holy trifecta of opportunity. So you've got a market transition, you actually have a chip ready for the market at the right time, and the number one incumbent, which is Intel, doesn't have a product. I mean, not that they're late, they just don't have a product. And that's the type of stuff that great companies are billed out of, are those unique opportunities in the market. And more than anything else, that's what brought me back to QLogic. All right, so before we dig in some of the ethernet and the hyperscale piece, what's the state of fiber channel, Sean? We said, is fiber channel the walking dead? Is it just a cash cow that QLogic will be able to milk and brocade and the others in the fiber channel business for a number of years? What's your real impression of fiber channel today? So look, fiber channel is mature, there's no question about it. Is it the walking dead? No, not by any stretch. And if it is the walking dead, man it produces a lot of cash. So I'll take that any day of the year. Hey, the walking dead's a real popular show, so fiber channel is still gonna be used in a lot of environments. But jokingly, the way that I describe it to people is I look at fiber channel now as the Swiss banking of networks. So a lot of web giants buy our fiber channel cards and people will look at me and go, why do they do that? Because for all the hype of open compute and all the hype of the front end processors and all the things that are happening, when you click on something where there's money involved, that's on backend oracle stuff and it's recorded on fiber channel. And if there's money involved, it's on fiber. And as long as there's money in the enterprise or in the cloud, I'm reasonably certain fiber channel will be around. Yeah, it's a funny story. I remember two years ago, I think we read Amazon's re-invent show and Andy Jassy's on stage and somebody asked, well, how much of Amazon is running, Amazon.com is running on AWS and it's most of it. And we all joke that somewhere in the back corner running the financials is a storage area network with the traditional array, probably attached by fiber channel. Absolutely, I mean, we just did a rollout with one of the web giants and there were six different locations. Each of the pods for the servers are about 5,000 servers. And as you would expect, about 3,000 on the front access servers, there's about 500 for PopCache, there was about maybe 12, 1300 for the big data and content distribution and all those other things. The last 500 servers look just like the enterprise. Dual 10 gigs, dual fiber channel cards and I don't see that changing anytime soon. All right, so let's talk a bit a little bit 25 gig ethernet. I had an interview yesterday with Melanox actually who had some strong claims about their market leadership in the greater than 10 gig space. So where are we with the standards, the adoption in Q-Logs position and 25 gig ethernet? Sure, so obviously like everyone in this business, we all know each other. And when you look at the post 10 gig market, 40 gigs been the dominant technology and I will tip my hat to Melanox, they've done well in that space. Now we're both at the same spot. So we have exactly the same opportunity in front of us. We're early to market on the 25, we have a race to get there and what we're seeing is the 10 gig market is going to 25 pretty straightforward because they like the single cable plant versus the quad cable plant. The people that are at 40 aren't going to 50, they're going to transition straight to 100. We're seeing 50 more as a blade architecture, mid-plane sort of solution. And that's where we're at right now. And I can tell you that we have multiple design win opportunities that we're in the midst of and we are slugging out with these guys everything and it will be an absolute knife fight between us and Melanox to see who comes out number one in this market. Obviously we both think we're going to win but at the end of the day I've placed my bet and I expect to win. All right, so Sean, can you lay out for us, where are those battles? So traditionally the network adapter, it was an OEM type solution. I got it into the traditional server guys and then it was getting the brand recognition for the enterprise customers and pushing that through. How much is that traditional kind of OEM? Is it changing? What's having the service providers and those hyper scale web giants? Yeah, so there's three fundamental things when you look at 25 gig, you got to deal with. So first off, the enterprise is going to be much later because they need the IEEE version that has backwards auto negotiation. So that's definitely a 17, 18, pearly transition type thing. The play right now is in the cloud and the service provider market where they're rolling out specific services and they're not as concerned about the backwards compatibility. So that's where we're seeing the strength in this. So they're all the names that you would expect and I have to say one of the interesting things about working with these guys is their NDAs are even nastier than our OEM NDAs. They do not want you talking about them. But it is very much that market where it's a non-traditional enterprise type of solution for the next 12, 18 months. And then as we roll into that next gen around the pearly architecture where we all have full auto negotiation, that's where you're going to see the enterprise start to kick in. Yeah, what are the types of applications that are driving this next bump in speed? What is it? Is it video? Is it sort of East and West types of application traffic? Is it big data? What's driving this next bump? So a couple of things you would expect which would be the certainly Hadoop, MapReduce, those sorts of things are going there. The beginning of migration to Spark where they're doing real-time analytics versus post-processing batch type stuff. So there, they really care about it. And this is where RDMA is also becoming very, very popular in it. The next area that most people probably don't think of is the telco and a v-space is the volume, as these guys are doing their double move and they're going from ATCA type platforms running mostly one in 10, they're going to leap right to 25. And for them, the big thing is the ability to partition the network and do that virtualization and be able to run DPDK in one set of partitions, standard storage on another set of partitions and classic IP on the third. Among the few folks that you would expect in that are the big content distribution guys. So one of the companies that I can mention is Netflix. So they've already been out, they're at 40 right now. And they're not waiting for 50, they're going to make another leap that goes forward. And they've been pretty public about those types of statements. If you look at some of the things that they talked about at IDF, and they're wanting to have NVME and direct DAS connection over ICER, that's driving the 100 gig stuff. We did a demo at a Flash Memory Summit with Samsung where we had a little over 3 million IOPS coming off of it. And again, it's not the raw number that matters, but it's that ability to scale and deal with that many concurrent sessions that are driving it. So those are the early applications and I don't think the applications will be a surprise because they're all the ones that have moved to 40. 10 wasn't enough, 40 might be too much, they're going to 25 and for a lot of the others. And it's really the pop cash side that's driving the 100 gig stuff. Because when that Super Bowl ad goes, you got to be able to take all that bandwidth at once. So Sean, you brought up NVME. Maybe can you discuss a little bit, what are NVME and some of these next generation architectures and what's the importance to the user? Sure, so NVME is basically a connection capability that used to run for hard drives and then as Intel moved into SSDs, they added this so you had very, very high performance, low latency PCI Express-like performance. What a number of us in this business are starting to do is then say, hey look, instead of using a SAS, which is kind of running at a gas at 12 gig, let's move to NVME and make it a fabric and encapsulate it. So there's three dynamics that help there. One is the advent of 2550-100. The second is the use of RDMA to get the latency that you want. And then the third is encapsulation of ICER or the iSCSI with RDMA together. And it's sort of that trifecta of things that are giving very, very high performance scale out on the back end. And again, this is for the absolute fastest applications where they want the lowest latency. There was an interesting survey that was done by University of Arizona on latency. And it said that if two people are talking and if you pause for more than a quarter of a second, that's when people change their body language. They lean forward, they tilt their head, they do whatever. And that's kind of the tolerance factor for latency on these things. And again, one of the statements that Facebook made publicly at their recent forum was that they will spend $100 million to save a millisecond. Because that's the type of investment that drives their revenue screen. The faster they get clicks, the faster they generate revenue. So when you think of high-frequency trading, when you think of all those things that are time sensitive, the human factor and that are going to drive this. All right, so storage in the interaction with networking is critically important, especially at a show like this, at VMworld. And I mean, Sean, you and I have talked for years, is it wasn't necessarily a fiber channel versus the ethernet, it's changing operational models. If I go use Salesforce, I don't think about my network anymore and if Salesforce happened to use ethernet, it's I don't really care. Hyperconvergence, when somebody buys hyperconvergence, you know, just kind of the network comes with it. When I buy a lot of these solutions, my networking decision is made for me and I haven't thought about it. So, you know, what's that trend that you're seeing? So for us, the biggest trend is that it's a shift in customer base. So people like Newtonix and these guys are becoming the drivers of what we do. And the OEMs are becoming much more distribution vehicles for these sorts of things than they are the creators of this content. So when we look at how we write and how we build these things, there's far more multi-threading in terms of them. There's far more partitions in terms of the environment because we never know when we get plugged into it what that is going to be. So incorporating our L2 and our RDMA into one set of engines so that you always have that high performance on tap, on demand. And you know, without getting down into the minutia of the implementation, it is a fundamental shift in how we look at our driver architectures. You know, looking at arm-based solutions and microservers versus just x86 as you roll the film forward. And it also means that as we look at our architectures they have to become much smaller and much lighter. So some of the things that we traditionally would have done in an offload environment we may do more in firmer on the side. And I think the other big trend that is going to drive that is this move towards FPGAs and some of the other things that are out there essentially acting as co-processors for this. You mentioned earlier Open Compute, Open Compute platform, those foundations and what's going on. What is, what's really going on there? I think a lot of us see the headlines. Sometimes you think about it and you go, okay, this is an opportunity for lots of engineering to contribute to things. But what's the reality that you're dealing with the web scale folks, they seem like the first immediate types of companies that would buy into this or use it. What's the reality of what's going on with that space? Well, obviously inside the, I will say the web scale, cloud giant space, I think right now, if you look at it, you've got sort of the Big Ten, Baidu, Tencent, Alibaba, Amazon Web, Azure, Microsoft being those guys. And then they are definitely building and designing their own stuff. There's another tier below that where you have the eBay's, the Twitters, the other sorts of folks that are in there. And they're just now starting that migration. If you look at the enterprise, not a big surprise, the financial guys are leading this. We've seen public statements from JPM and other folks that have been at these events. So I view it very much like the Blade server migration. I think it's going to be 25% of the overall market, whether people like to admit it or not, good old rack and stack is going to be around for a very long time. And there are applications where it makes a lot of sense when you're deploying private cloud in the managed service provider market, we're starting to see a move into that. But if you say, what's the 10-year life cycle of an architecture, I would say that in the cloud, we're probably four or five years into it. In the enterprise, we're maybe one or two years into it. All right, so what about the whole SDN discussion, Sean, how much does Q-Logic play into that? What are you seeing in general and we're at VMworld. So what about NSX, is that part of the conversation and what do you hear in the marketplace today? Yeah, it really is part of the conversation. And the interesting part is that I think SDN is getting a lot of play because of the capabilities that people want. And again, when you look at the managed service providers wanting to have large scale, lower cost, that's going to definitely drive it. But much like OpenStack and Linux and some of these other things, it's not going to be, you know, the guy's going to go download it off the web and put it in production at AT&T. You know, it's going to be a pre-packaged solution. It's going to be embedded as part of it. You know, if you look at what Red Hat is doing with their OpenStack release, if you look what Mirantis is doing with their OpenStack release, again, from an enterprise perspective and from a production in the MSP and second tier cloud, that's what you're going to see more of. So for us, SDN is critical because it allows us to then start to do things that we want to do for high performance storage. It allows us to change the value proposition in terms of, if you look at Hadoop, one of the things we want to be able to do is take the storage engine module and run that on our card with our embedded V-switch and our next gen chip so that we can do zero stack copies between nodes to improve latency. So it's not just having our DMA, it's having a smart stack that goes with it. And having the SDN capability to go out, tell the controller, pay no attention to this little traffic that's going on over here. These are not the droids you're looking for and then everything goes along pretty well. So it's very fundamental and strategic but it's a game, it's a marketing which we're going to participate but it's not one we're going to try and write or do a distribution for. Okay, any other VMware related activities, Q-Logix doing announcements this week that you want to share? This week I would have to say no. I think the one other thing that we're strategically working on with that you would expect is our DMA capabilities across V-Motion, V-SAN, those sorts of things. We've been one of the leaders in terms of doing Genevieve which is the follow on to VxLan for hybrid cloud and that sort of thing. And we see that as a key fundamental partnership technology with VMware going forward. All right, so let's turn back to Q-Logix for a second. So the CEO recently left. He did. There's a search going on so give us the company update if you will. Well actually, there isn't a search. So Genevue is going to run the ship forward as CEO. We've brought in Chris King who was on our board as executive chairperson. Chris has a lot of experience in the chip market and she understands that intimate tie that we have to that Intel TikTok model and really how you run an efficient chip driven organization. Whether we play in the systems in between level where we're not quite the system but we're not quite the chip and understanding that market is part of what she does. And the board has given us the green light to continue to go forward, develop what we need to do in terms of the other pieces. Gene has a strong financial background. She was acting CEO for a year between HK and Simon or after Simon left. So she's got the depth, she knows the business. And for us, it's kind of a non-op where everything else is continuing on as you would expect. Okay, last question I have for you, Sean. I mean, the dynamics chain for years, you know, there was kind of the duopolys in the market. I mean, it was Intel and Broadcom on the Ethernet side. It was Emulex and QLogic. It's a different conversation today. I mean, you mentioned Intel, we talked about Mellanox, there's QLogic, you know, your old friends and Emulex, Avago bought Broadcom and now they're called Broadcom, I think. So, you know, layoff for us, you know, kind of, you know, where you see the horses on the track and, you know, what excites you? Yeah. So again, you know, if you look at the 10 gig side of the business, clearly Intel has the leadership position right now. We're number two in the market. If you look at the share data that's come out, you know, the Emulex part of Avago has been struggling and losing share. Then we have this 25 gig transition that came in the market and that was driven by Broadcom and, you know, for those of us who have followed this business, they, I think everyone can appreciate the irony of Avago buying Emulex and then for all the years, we tried to keep them separate, bringing them back together was, we've chuckled over a few beers on that one. But then you've got this 25 gig transition and, you know, the other thing is that if you look at, so let me step back and say the other thing on the 10 gig market is it was a very, very clear dividing line. The enterprise was owned by the Broadcom slash QLogic, Emulex side, the cloud, the channel, the appliance business was owned by Intel Melanox. Okay, now as we go into this next generation, you've got us, Melanox and the original Broadcom team coming in with 25 gig. We've all done something that gets us through this consortium approach. We're all going to have an IEEE approach from there and Intel isn't there. You know, we haven't seen any announcements or anything specific from Emulex that they've said publicly in that space. So right now, we kind of view it as a two horse race. We think from a software perspective that our friends at Broadcom, Broadcom, whatever we want to call them or Bravago, I think is how our CTO refers to them, that I don't think they have the software depth to run this playbook right now. And then what we have to do is take our enterprise strength and move those things like load balancing and failover and the SDN tools and N-par and all the virtualization capabilities we have, we got to move those rapidly into the cloud space and go after it. For us, it means we have to be more open source driven than we have been in the past. It means that we have a different street fight for every one of these. It represents a change in some of the sales model and how we go to market. So not to say that we've got all the everything wrapped up and perfect in this market, but again, right time, right place and this will be the transition for another, we think three to five years. And there's still a lot of interesting things that are happening. Ironically, one of the most interesting things I think that's going to happen in 25 is this use of the new little profile connectors. I think that will do more to help the adoption of 25 gig and 100 gig, where you can use the RxC connector. RxC, I forget the acronym, but it kind of looks like the FireWire HDMI connectors that you have in your laptops now and now imagine that you can have a car that has that connector in a form factor that's maybe a half inch square and now you've got incredible port density and you can dynamically change between 25, 50 and 100 on the fly. Well, Sean, we've always talked, there's a lot of complexity that goes in under the covers and the industry's got a good job of making that simple and consumable and help drive those new textures forward. All right, Sean, thank you so much for joining us. We'll be right back with lots more coverage including some more networking in-depth conversation. Thank you for watching. Thanks for having me.