 and it's ecosystem sponsors. And now your host, Dave Vellante. Back to San Francisco everybody. I'm here with Stu Miniman. Stu, first time we worked together this week. I'm excited to have you with me. We're here at Moscone. VMworld 2015, our sixth year. Kevin Dierling is here. He's the Vice President of Marketing at Melanox. I love Melanox. I love Infiniband. I love High Speedy. It's all good because it's all about getting data from here to there. And that's what you guys are really good at. And moving data is hard. Absolutely. So, how do you feel? What's going on? What's new? It's great to be here, thanks to be on theCUBE. Great to be here in San Francisco at Moscone Center at VMworld first time. First time in theCUBE. First time in theCUBE. CUBE newbie, fantastic. Well, welcome. Excited about that and excited about moving data fast. Because as you said, it's all about faster storage and faster storage needs faster networks. And so when we look at the software defined data center and we think all of the software, it's really about building a data center where all of that data movement is being done in software. And what we do is make it much more efficient. We make it much more efficient to move the data and to run the applications because the amount of data that's being produced now, big data applications, there's simply no way to store it on a single computer. And the key concept is, is if you can look at the data and not care whether it's on your server, inside of some flash, or it's something on another server and bring it over very, very quickly, that's what this is all about. So things like vSAN and Cep and all of these scale out storage platforms that we're seeing, we're running on Scale.io showing in our booth, we get two and a half times more virtual desktops when we run on fast interconnects. So that's the real key about moving data quickly. Yeah, and I mean, distributed computing has always been a huge challenge. Scale out's been a huge challenge. And now you see this, we all know about data growth. It's just, it's massive. We all talk about it. When we start thinking about, well, you mean you see things like Hadoop and everybody talks about these nodes all over the place and you start projecting these infrastructures that are going to support that and people say, okay, but how do I move the data? How do I recover? So that's where you guys come in. Exactly. So if you look at some of these Hadoop and where that was originated in some of the hyperscale cloud, that's exactly where we play. We're actually the dominant supplier of Ethernet NICS into that hyperscale platform. So we have over 85, 86% market share at anything above 10 gigabit Ethernet. So at 40 gigabit Ethernet today, we are the leader and people don't realize that and it's all about moving the data. Those are the guys, some of these guys have 160 petabyte data stores that they go search. And you don't realize when you do something as simple as searching that you're doing massive East West traffic inside the data center. And so there's only a tiny little bit of traffic that's going to your phone, but all of your search history and everything that you've done over the past months is going in and your location-based services, it's figuring out where you are. There's a thousand times more traffic inside of the data center than there is between you and the client. Kevin, why is it so hard? Why is Melanox so successful? Is it because others just aren't focused on that? Is it your secret sauce? I wonder if we could talk about that for our audience. I think it's the secret sauce. We have really a heritage with Infiniband of building very, very large computers, 20,000 nodes with high performance computing. That's our heritage where we came from. We sort of started with things like SDN and scale out software, scale out file systems. That's where we started. And about five or six years ago, we entered the Ethernet space and we're really taking that heritage and bringing it to Ethernet. So we have far and away the most efficient Ethernet technology in the world. We have something called Rocky, which is RDMA over converged Ethernet. And it lets you move data from one machine to another machine without involving the CPU or the operating system. And that's what's so critical. We can get much higher performance. We showed at Microsoft, I shouldn't say the Microsoft word here at VMworld, but there you go, everything's fair game. So, but we showed at their event, the Ignite they showed on our 100 gigabit Ethernet adapters, they got line rate. So they got 94, 95 gigabits per second, actual throughput to storage using just three flash NVMe drives inside the device. And when they turned off our secret sauce, the Rocky, and they just tried to use regular software transport, the throughput got cut in half. Okay, so they only got less than half of it. And then they showed why the CPU was completely pegged. They had four of the cores running, trying to move the data back and forth. And they just couldn't keep up. They ran out of CPU power. And you think about it, you buy an expensive server to run an application, not to move the data around. So the key thing is, you don't want to be spending all of your CPU. That's the most expensive part of the entire server. You don't want to use that up just to move the data. You probably have an application, whether it's Hadoop or some analytics on top of that. You don't want to just be moving the data. You want to be analyzing the data and creating something you can monetize for your business. So that's the real value of what Melonix brings. So Kevin, our CTO at Wikibon, David Floyer has actually written a lot about how the HPC architectures are going to be pulled into more traditional environments. When Flash came out, we knew that it was going to move that bottleneck. Because of course the biggest challenge we always have is you're never eliminating bottlenecks, you're getting rid of it. So storage hadn't been moving fast enough. So now Flash, it changes that bottleneck and we need the transport layers. So the fiber channel guys have been saying, you want low latency, you want good bandwidth. I mean, fiber channel does a good job there. Infiniband has a great legacy there too. So bring us up to speed is where we are from the Ethernet standpoint on kind of that comparison. Yeah, so the fiber channel guys like to talk about that, but in fact there is no fiber channel in the cloud. There isn't any fiber channel. We're selling Ethernet into the cloud and we're selling Infiniband. So there's a lot of these big clouds that are actually running Infiniband inside. You don't even know it. But really, we just don't see fiber channel. We're not competing against it. There was new data that just came out that showed that fiber channel is declining. I know you wrote a piece that said that the market for server sand, what Wikibon talks about and what David's written about is growing at 38% you know, cagger. And meanwhile you've got traditional sand declining at something like an 8% and we saw new data today that confirms that. So from our perspective we're very much aligned with Wikibon saying that scale out storage with the server sand, that's what's happening and it's on Ethernet. It's interesting because if you look at say the all flash array market in the enterprise data center, there's some fiber channel there but I think you're differentiating and saying okay, if I look at the largest companies in the world, when I'm doing things like Siri or all those other hyperscale type environments, yeah that's not running on fiber channel. But it's never been a matter of if it's when, right? I mean you've always felt that way. It just takes a long time. Yeah, great point Dave because if you talk to networking people, it's the Nirvana has always been I want to get down to a network. I mean that's what Infiniband got created to do is let's start from the ground up but it's always been kind of when Ethernet, not if Ethernet, it's been what most of us have thought for gosh, I mean the last 25 years. So we certainly, we're an Infiniband company only. At one point that was 90% of our business today if you look at it more and more, Ethernet is what's driving our business. It's certainly driving much of the growth. We dominate the HPC space, we have some pretty significant market share there but now we're growing much faster in storage and cloud and web 2.0 and these hyperscale businesses and what's exciting for us is we see some of the hyperscale and the cloud activities are going to move down into private cloud and enterprise activities and we see some growth possibilities there that really haven't happened before. Yeah, so Kevin, I'm wondering if you can give us an update on just kind of the speeds and feeds discussion because Ethernet used to be real simple. You know, like every decade we did a 10x improvement and it took us a bunch of years to do it and we kind of got bogged down. I mean, 10 gigabit Ethernet ratified back in 2002 has taken us a long time. If I talk about the enterprise, it's still only about, I don't know, a quarter to a third of all servers are shipping with 10 gig. But you're pushing beyond 10 gig, it's not just 10, 40 and 100 but now there's 25 and there's 50 so just give us some of the update on that. Sure, yeah, it's actually kind of exciting times when you see some of these changes that are happening. We really think that 25 is the new 10. So if you look at it, it uses the exact same fiber infrastructure that you can aggregate across existing fiber plant. So you can plug in an LC connector, just use a new transceiver or a new adapter in our top of rack switches. So we're shipping end to end. We've got adapters, we have cables and we have switches. You can aggregate with 100 gigabit. So whether you want to use 25 gigabit down to the servers or 50 gigabit and then you can aggregate between switches for your scale out at 100 gigabits per second using the same infrastructure that you've already installed. So that's pretty critical. So really when you look at it, if you're getting two and a half times the bandwidth of 10 gig, okay, why would you use 10 gig anymore? So the only reason would be price. We're seeing the price of 25 gig come down very, very close. And if you look, I don't know the history of this, really the industry is driving this. So at first there was a call in the IEEE to standardize 25 gigabit ethernet. And the IEEE is a very great organization but it sometimes is driven not only by technical considerations. There's other considerations that drive it. And in fact, they said, no thanks. We don't want to standardize it. So it took ourselves, Melanox and then Microsoft, Google, Broadcom and Arista came together. We formed the 25 gigabit ethernet consortium and we said, we're going to make this happen and we're going to have interoperability. That's happened since then. We've seen a lot of other companies, Cisco and Dell have joined at the top members of the 25 gig ethernet consortium. And lo and behold, the IEEE has now embraced it as well and they're actually voting to standardize it. So we think 25 gig is here, worshipping end-to-end solutions. We think we're the only vendor that has end-to-end 25 and 100. So you said before, at one point, you were kind of a one unifocus company on Infiniband and Infiniband was able to do things that ethernet wasn't. Is that, am I understanding that gap is closing? I mean, we've seen what Oracle has done with, I mean, Larry Ellison loves Infiniband. I've written about it and for good reason. Are we seeing that gap close now or the Infiniband business continue to chug along? I don't think it's really so much that the gap is closed. There's still, if you want to build the most efficient, the most high-performance communication, storage, compute platform in the world, it's based on Infiniband. So if you look at the high-end of EMC's, VMAX platform, there's Infiniband inside of that. Really, if you look at any of the enterprise platforms, so the Exa platforms from Oracle, if you look at Teradata's, Hadoop platforms, it's all running on Infiniband. So Infiniband is the most efficient network and platform in the world to build scale-out systems. That said, there's a whole class of people that really are going to use Ethernet. There's what Metcalf said, Ethernet and Ethernot. And so for those folks, we said, why should we fight this anymore? There's a huge market. We're really good at building networking equipment, switching infrastructure, and most of what we can do is actually common. So we took what we had, this heritage that we had, and that's what Rocky is. If you look at Rocky is, we brought REMA to Ethernet and pioneered that, opened it up to the industry. Now we see a Rocky alliance that forms an initiative within the trade association. We see lots of other companies now that have joined that and are offering solutions. Part of the value here of Ethernet is that you have multi-vendor interoperability. And so we're seeing that happen with Rocky. It's not always an equal, you know, protocols, switching protocols is not always an easy thing. You saw a lot of the fiber channel guys try to get it to Ethernet or even Infiniband and failed. And you're saying the success is due to Rocky. Yeah, I think that was. That IP. That was part of what really drove it. The other thing is, you know, sometimes when you're entering a new space that maybe you're an adapter company and you're trying to enter the switch space, okay? Those are really different disciplines and it's pretty difficult to actually bridge from an adapter to a switch. Melanox was always an end to end company. We've been an end to end company. We've been doing switches. We've been doing adapters. And they work together very, very nicely. So today we think we're the only company out there that's really shipping 100 gigabit products end to end. And so that's exciting. All right, so Kevin, can you talk a little bit about the update on VMware solutions? And it's okay to talk about Microsoft too, but kind of the virtualization space, what are you seeing from a networking standpoint? Yeah, so what we've done is really building on some of the things that VMware themselves did. So Evo and Visan. So we've taken our own reference architecture to bring that to higher speeds. So 25, 50, 100 gigabits in our booth. We're actually showing, I think it's about 96 gigabits per second with VMware on Visphere. So this is something that we can qualify ourselves. You know, ultimately we'll work with our partners to get that out in the market. Same thing with Visan. So we're seeing great scale out performance, as you can imagine. You know, we have to work through all of the software. Oftentimes people, they love and they hate us because we in fact show the weaknesses of their software. People will come and use our networks because we have so much performance that suddenly we can expose it was what you were talking about. There's a new bottleneck. Used to be that spinning rust that was the bottleneck. Now Flash, okay, has taken that away and the network becomes the new bottleneck. We've taken that away and now all of a sudden you've got to find a bottleneck in the software. So sometimes people love us, sometimes people hate us. We work very closely together and we're always leapfrogging the technology. Excellent, so now, what about the event here? What do you got going across the street? You got a booth? We got a booth. What's the action like? Yeah, it's great. It's fantastic. We've got a bunch of partners that are presenting there. So almost all of the storage vendors are showing our networking equipment to show how they scale out storage. There's things like NVME over Flash, which is a new technology which is being developed. We're showing the vSAN technology. We're showing a CloudX, which is our reference architecture running vSphere and it's running all of our advanced Rocky features for storage. So we can do iSCSI, for example, over RDMA. So we run that over Rocky. We get about two and a half times the number of virtual desktops. We're showing that with Scale.io in our booth. We're showing it with vSAN. So we have a lot of different flexibilities. We work across a lot of different platforms and so a lot of people are coming by to see. There's not too many 100 gig switches in the world. There's not too many 25 gig adapters and switches. People are coming over and they're excited to see it. It's obviously a great event, huge, I don't know, whatever, the number is 20,000. 23,000 plus, yeah. So that's pretty, that's up, actually, if I think from last year. It was 22,000 plus last year. Yeah, okay, so that's cool. Huge show, tons of customers, but we also saw you at Red Hat, right? Absolutely. Interesting crowd, DevOps crowd, you know? Yeah, so the Red Hat was interesting. That was really more of an OpenStack story. So there you've got the OpenStack, which is the folks that are really focused on, obviously, open source. For us, we actually do all three. So we're sort of agnostic, when I say all three, all three of the major cloud platforms. So you've got the obviously VMware and with vSphere. We've got the Azure platform, that publicly, Microsoft has talked about the fact they're using us inside of their Azure platform and now they're taking that into the private cloud. And then we have OpenStack and we have reference architectures for all of those and we work with partners. We really get much more efficient application level efficiency. So that's the key. When you mentioned Arista before, they're a partner, right? And of course. They're more of a competitor. They're more of a competitor. But also, but a partner in the sense of the consortium. Absolutely. And then you compete. It's a competition we live in. Absolutely, absolutely. It's an interesting world. Well, Kevin, I'll give you the last word. Sort of bumper sticker on VMworld 2015. What would you summarize what you've seen so far? I know it's only day one. Yeah, it's day one. It's a great show. We're really excited. People ask us, what is the storage show anymore? An EMC world for sure, but really VMworld as well. So we're excited to see all the storage activity here. We think that faster storage needs faster networks. 25 is the new 10. There's no fiber channel in the cloud. So we've got a great opportunity here. We think we're talking to a lot of partners and we're looking forward to having some good business. Awesome. Moving that data, it's just going to get bigger and bigger. So it needs to be faster. Kevin, thanks very much coming to theCUBE. Absolutely. Thank you. All right, keep right there, buddy. We'll be back to wrap up VMworld 2015 day one just after this brief word right back.