 Live from Las Vegas, Nevada, it's the Cube at HP Discover 2014, brought to you by HP. Good morning from Las Vegas everybody, this is Dave Vellante and this is the Cube. The Cube is Silicon Angles flagship production, we're here live at HP Discover. We go out to events like HP Discover and many others, we extract the signal from the noise. We've been passing a lot on the server business last night, we saw Antonio Neary unveil a new system from HP, a new water cooled system that we're going to talk about a little bit. I'm very excited because I've seen water cooled go full circle from the, well predated even my entrance into the computer industry but now it's back and what you're going to see is very interesting. Alan Andreoli is here, he's the Vice President and General Manager of the Density Optimized and Service Provider and HPC Business at HP and John Gromala, John we were talking off camera I think it's the 7th time, maybe 8th time you've been on the Cube, so welcome gentlemen. Thanks, good to see you guys. So Alan I want to start with you, talk a little bit about this group that you run, it's the Hyper Scale Group essentially, talk about its mission and how you're going after that market. There is a revolution taking place in the world of computing, it's Hyper Scale, so you are moving from traditional enterprise applications to more and more service providers, enterprise high performance computing and then the private cloud in the enterprise, this is driving a new generation of data centers, so the organization that we have created concentrates around the key aspects of this transformation. We have a product line that is density optimized computing that really tackles with the needed higher density for the compute and then we have two vertical markets, one which is service providers for the cloud, for the web and one which is high performance computing and these two segments are the fastest growing ones in the whole market and they are together moving from one side of the market today to over half of the market in the next two to three years and this is what our group is leading for HP servers. So Hyper Scale, we talk a lot about Hyper Scale on the cube, we've noticed that a lot of the Hyper Scale trends are starting to bleed into the enterprise, in fact some folks have said if you want to know what's going to happen to the enterprise, look what's happening inside of these giant Hyper Scale companies if you can peek inside and parts of it will seep into that enterprise, do you buy that premise? It's a kind of a circle if you wish, we're talking about a circle for water cooling, we can talk about water cooling later on, but it's the same for the Hyper Scale momentum. It started with the largest providers of the cloud, all who provide very very optimized and unique monolithic workloads, the Google's, the Amazon, the Facebook, the Microsoft and then the ones a little smaller, what we call the tier two service providers are looking at what they are doing and then the tier three's are looking at the tier twos and then the enterprises who are also building Hyper Scale clusters are looking at what the service providers are doing. So you have this kind of pervasion in the market of the Hyper Scale happening that starts from the largest Hyper Scale providers, the service providers. Yeah, so that's a good way to look at it. So the tier one guys, a lot of people say it's a race to zero, I personally don't think it is, it's massive volumes, huge innovation and then the tier two and the tier three service providers in Hyper Scale are competing with the Amazons and the Googles by specializing in areas where they may not and then enterprises, some of the enterprises, particularly financial services, are looking at what's going on in that Hyper Scale site, saying maybe we can drive similar efficiencies. It's a very interesting market dynamic. Specifically you've got a situation where the true, the large scale Hyper Scale guys, they'll spend a lot of PhD time to save money over the long term, whereas the average enterprise doesn't have that capability, they'll spend money to save time. So there are clearly differences, but one of the things that we notice, maybe let's say five or six or seven years ago, conventional wisdom said that the Hyper Scale guys will just use commodity components and have the software defined, what's now called software defined, on top of it, high degrees of automation and now you're seeing that change a little bit. They're going highly customized. So what will happen in the enterprise? Will the enterprise go in that similar cycle? Commodity, software defined, that's what you're hearing now. A lot of talk about it, and then eventually come back to highly customized. HP knows a lot about customization. What are your thoughts on that? It's all about segmentation. The answer is yes and yes. So I think that we are really cracking the code here at HP of what the market needs to be successful in this transformation. On one hand you have the people who want to go to customization and ultra commoditization of the architecture around their workload. And that's why we have done this disruptive John Venture with Foxconn. What we're going to be able to provide is bare metal, customized to the needs of the biggest service providers on the planet and beyond. So for the people who go for customization of the hardware for massive workloads, we will be the vendor with no way, how scale can be matched end to end. The HP brand and the power of Foxconn behind in terms of supply chain and chip engineering. For the enterprise side, there is a need for innovation. Some of them want to use bare metal, but some of them want some unique solutions to do. Compute solution with ultra performance, high density, and this is the Apollo product line. So we are going on the two fronts with a very differentiated approach and basically enterprises can choose which way they want to go, depending on the level of capabilities that they want to build for the future. We talk about it in terms of workload optimized and there are differences in those workloads between enterprise and those largest tier one, tier two customers. So especially the enterprises, they want the benefits of those scale out systems that hyper scale. But they still have some different requirements and manageability and the enablement of things like OpenStack and where that's going, are key enablers that will take place over the next few years as well. Well, that's a great point because the average age of the enterprise application is probably 15 to 20 years. Absolutely. The average, whereas a Google or a Facebook, I mean a lot of people say they have one app, maybe one main app and a lot of little apps around it, but a whole different paradigm, isn't it? So how do you, what parts of the hyper scale service that enterprise sort of application portfolio? And what parts are new maybe around the edge that maybe they're developing mobile apps? I wonder if you could talk about that dynamic a little bit. Well, it's even a case, let's take HPC as an example. HPC used to be something that was more government and academic in terms of where it was prominent. And it has really started expanding where it's a key tool for enterprises now. And that has all those same hyper scale characteristics. It's a key piece of how you deal with those things. So we can take all of those benefits and apply it to how engineering, technical folks do their core work and what they're doing with a lot of the new platforms that we've built. And a lot of the big data stuff in HPC are kind of colliding as well. Correct, exactly. Both of those are benefits. So let's talk about some exciting stuff. I talked about water cool before. Dust off your thermodynamics, 101 engineers out there. You'll love this segment. So talk about inside of what you announced last night or unveiled last night. You've got a new system that's water cool. But we think water cool, we think chillers, right? We think frost on the outside of the pipes. Talk about the innovation that you guys have developed inside of Apollo. Go ahead, Johan. Yeah, well, so it's been four years in the making. We've been looking at the HPC market coming for potential explosion, like a second life for HPC. As John was saying, it used to be pretty specialized in academia and research and so on, government. And now it's going mainstream. So we said if this is going mainstream, it's got to be really disruptive in terms of technology and innovation. And the more power you put, the hotter it gets. So cooling is actually the first problem we've been trying to address. And you can see here our dry disconnect technology, which is unique. And I can maybe briefly explain the issue with water cooling is that you have risk of leakages on the electronics. With this technology, the water is going in the bus in the chassis, totally separated from electronics. And then you have a liquid here, which goes into vapor with the heats, get cooled down back into liquid here on the bus, and then back into vapor, it gets into the heat. And we have achieved an extreme efficiency with a ratio of 1.06 of energy utilization, which is totally unmatched. So that's the first innovation. Yeah, 1.0 is nirvana. 1.0 is possible to achieve, but it's perfect. One watt in, one watt out, impossible. So 1.06 means extreme efficiency. That's what we have achieved. Unless you're including the reuse though, you could actually go with PUs that are lower by having the reuse of the technology of the waste heat as well. So you get even greater efficiencies when you reuse a lot of the heat from that liquid cooling as well. Now, we were talking last night, this is not chilled water. How did you achieve that innovation? Yeah, again, it's about a warm water cooling. So you take air temperature water and actually put that in the system. Since the actual chips are running at a much higher temperature, you could use that difference in that temperature alone to go ahead and cool the system. And that is adequate to deliver the reliability and availability that you need. Absolutely, and different customers and vendors will actually decide whether they want to do just strict cooling of that or whether they want to reuse that heat. There's so many benefits in taking that warm water and reusing it to heat facilities or actually doing a lot of offsetting of energy that's used for other purposes. Where did this innovation come from? Was it inside of the server division R&D? Was it a collaboration with HP Labs? Can you talk about that a little bit? Both. It started from HP Labs who invented this kind of Dry Disconnect methodology. Yes, please. Okay. And then we productionize it with HP servers, making it an overall system architecture. Okay, so the ambient water comes in here and it's cool enough relative to the electronics. Yes, absolutely. And this liquid is kind of magic. You can kind of simulate, imagine that you have, and you can see these on the booth, imagine that you have an ice with glass and heat. You put one of these tubes, imagine these tubes would be straight. You put one of these tubes in the ice, immediately you feel the ice on your finger. You put a normal piece of metal, it will take a few seconds to go up and you will have a lot of dissipation of the cold. If you do it in a hot pot of water, you will have the same result very, very quickly. This is really breakthrough. And the material here is what? This is copper. It's copper, okay. But it's got a vacuum internal to it with the liquid, so you'll actually get to the point where that liquid will boil. You want to call it that change to vapor? And by changing from liquid to vapor, that actually uses a lot of energy and transfers it very effectively across that system. I want to ask you guys, short on time, but the OCP initiative out of Facebook, is that something that you guys are looking at, something that you're actually actively investing in, Yes, we are participating in the OCP movement. We have discussions and meetings and products available. We work with Rackspace, we work with Facebook, and we are trying to find ways where some segments of the market, probably in particular financial services, we find benefits in working with this type of architecture. Customers need to look at the overall cost of energy, total cost of energy, not individually the price of the product, but our John Venture with Foxconn will be supporting this type of generic services architecture. That's an OCP like architecture. OCP, WCS, and this type of. Yeah, yeah, okay, great. I think it's pure of a bigger trend towards overall open design principles. And we even have a session here at the event on OCP and those open design principles. Last question is how big, so if the enterprise market is X, I don't know, let's say X is 100 billion, what percent of that market is your, do your solutions, do you think attack? Is it a? 36 percent. That's your number. This year. That's your TAM. This year, yes. Okay, that's a TAM figure. Or the total market. 36 percent, that's a TAM. Your total available market is 36 percent of the overall enterprise market. That's a today statement though, isn't it a lot of work? That's a today statement. Today. So over time it's obviously going to increase quite dramatically. We believe it will pass 50 percent in 2017. So this is the future. You believe? I believe it too, by the way. I think these, I have one of the people who said, you want to know what's going to happen in the enterprise, look what's happening in hyperscale. And we will continue to have a very nice business and in the enterprise because traditional applications will continue and then there is a SME market which is not going to go hyperscale. Right. So you have SME, then you have traditional apps in the large enterprise, then you have HPC and these are different segments that you have to segment by workloads to understand where this is all going. But where the SMB market will go hyperscale is through the tier two, tier three cloud service providers, right, that will hop on this technology. Eventually, but not totally. Not tomorrow. Not totally and not everywhere. Excellent, that's a great segment. Thank you guys so much for coming on. We got to leave it there. Thank you Dave. Thank you very much. Thank you. All right, keep it right there. We'll be right back with our next guest. This is theCUBE, we're live from Las Vegas. Right back.