 Okay, we're back inside theCUBE. This is our flagship telecast. We go out to the events and extract the signal from the noise and share the knowledge of our guests with you. I'm John Furrier, the founder of silkenangle.com, silkenangle.tv, and I'm joined with my co-host. I'm Dave Vellante of wikibon.org. We're very pleased to invite back Pauline Nist, who's a CUBE alum. Pauline, welcome to theCUBE. From Intel, general manager of the server division. We really appreciate your spending time coming back to theCUBE here at HPE Discover. You guys have had a busy week. I've been following you online and you're all over the place this week. You know, it's our summer chores, not even the summer yet. Yeah. The demand for theCUBE is great and we're going to launch a 24-7 channel in the fall. Oh, cool. We're going to have morning and afternoon anchors and then all-perhing. So I have insomnia. I can watch theCUBE. There you go. Well, I mean, there's so much distribution. Xbox. Yeah, yeah. JustinTV, all the outlets, there's so much demand for the content because you can't get this on cable. Well, you know what you're going to need now. You're going to need a big data installation to get all of that social media data about who's watching you and where they're watching you and what the hotspot topics are, you know? Careful, John's going to show you our demo. You see it. The demo's expanding every day. As you know, we showed you, we use big data to do predictive analytics for our media business and we just spun that off as a separate company. Oh, cool. I hadn't heard that. We have a data scientist and developers on board. Hadoop, it's really a cutting-edge age base. And I do owe you coming to see that. We're going to show it to you. We want you to look at it because you're so smart, we want to get your perspective because we're learning. We're all learning. We're trying to figure it out and we're learning some gradient sites, but more importantly, we're not going to try to force it. We're going to let it kind of flow and see how this thing develops, but it's exciting. We're seeing things and we're meeting people we've never met before with these tools because now we have access to long-tail of people. And it's just so exciting. Well, I actually think tools like yours are almost part of what's going to displace the search engine mentality. I know that laying in front of the Google train is a very dangerous thing to do and I could be in like six pieces, but they're not breaking. They're not hot spot of the minute. I was telling somebody the story this morning of when the plane went down in the Hudson. My niece had a good friend who was in a building facing the Hudson River and saw that plane. I mean, I had pictures of that plane in the river before anybody was reporting it anywhere and that's the way media travels today. I mean, that's the way information comes about. What do you, okay, since you're here I'll just show you the demo. So we can search, this is our 2.5 million users. So we have now a search function. So I can get all the Hadoop affinity of all the, whoops, did I log off? Yeah, it logged off. So we now in real time can see what people are doing in our communities. Globally, by the way, this is not just our site because we have site analytics, but this is like affinity ranking. This is all the Hadoop affinity, people that have affinity towards big data and specifically Hadoop classified. So just random, we just pick one here, just out of the blue. Founder, still startup. Oh, they're not stealth anymore. He's the CEO, so all of the VMware, big data, cloud, serial entrepreneurs. So I would never, these people are all out there and that was real time by the way. So we can monitor them, we see the conversations and what that does is allows us to get new insights. What are they talking about? What are they sharing with us? So Hadoop has enabled us to do that. So my question to you, as you mentioned before we started, what do you see with Hadoop and around some of the big data applications? Because Hadoop's been great for us. It's batch kind of going to near real time. It's not really near real time yet, but it's, you know, seconds is, I guess near real time. Real time is defined milliseconds. But what do you see there? Is it's going to be cores required, flashes required? Oh, we love it. We love it. More cores, more streams, you know, more sockets, it's a great product for Intel. I just think that it's taken information, as you say, to a whole other tier in this near real time. Because I certainly understand that there are people who want absolute real time. If you're instrumenting an oil well or a production line or something, you know, if you want to monitor and adjust equipment, you actually need real time. But I think for a lot of the media, retail, you know, even healthcare kind of stuff, near real time is going to be fine and it's going to give people an order of magnitude over what they have today with batch. I mean, I was talking to Macy's people. Macy's goal is to unify their web and their store inventories and to be able to reprice all inventory in under an hour, which today takes them days, if not weeks. And that's an incredibly powerful thing, particularly when you're selling on the web. They're also going to unify the inventory so that if today, if they're out of something on the web, you're out of it and your only choice is to go to a store and see it, you can't see the store inventory. They're going to change that and they're going to equip their biggest stores to be able to direct ship from you and literally buy like it was web inventory, but it's in the store and you can get it. And that gives them much better pricing control because today, if something sells out on the web but doesn't move in the store, they market that on the store. They don't ship it over to the website and sell it for full price. But I'm where you are, which is the other thing that's really exciting is just following what's happening, what's changing, what's emerging, what people are doing. If you're in the media business, that's what it's all about and I'm going to have to get you to sign me up as like a beta site or something so I can look at this whole stuff. Awesome. We'll do it. We'll do it as well. I mean, first of all, we love having you on. I mean, look at it. We want you to look at it because we need your help. But I wanted to ask you about the near real time because I think real time scares people because not everyone's doing an oil rig or doing hedging and financial and or other high capacity, I mean, high performance. So near real time is good. And but let's change it to more of an architectural question. Let's take companies like SAP, application workloads and it's weeks to minutes. So what is it in the infrastructure in terms of the components now? How's that changing? We're seeing, for example, I was just talking to a Flash provider off the record and they say they take an SAP workload, a normal application on the web, five minutes to load to five seconds with Hannah and Flash. So these are the kinds of architectural changes that are changing the experiences of consumers and customers. What's your perspective on the system architecture here? What's your view? I think two things are happening. I think that people are learning a lot more because of the capabilities are available today about how to manipulate the data. I mean, everybody's doing, it used to be kind of lunatic fringe, but everybody's doing compression and decompression now. I mean, the way you make this stuff real time is you want to keep it in memory. The way you keep it in memory is you got to have a smaller footprint because today, if you want terabytes of memory, it's still pretty, you're going to spend hundreds of thousands of dollars to do that. So you've got compression, you've got, and that's being enabled by computes. I mean, it's nothing more than compute cycles to compress it and decompress it. The same thing with the Flash stuff, and I think we're on the cusp of seeing in the next three years, four years, another major revolution in memory technology. We talked about this last time, I think next generation NVRAM is going to be denser, it's going to be different technology. You've got all the big guys playing around with it in the lab today, certainly not ready for being a product. I mean, HP published a major paper on their Memrister Technology last November at a European conference, which I blogged about if you want to go read it. And I think that what it means is that you're going to continue to see as these new technologies start to show up, the revolution continue. I mean, we're investing in our many integrated core chip, which is a compute chip for high performance computing, but I think you're going to see data people look at it for driving decompression, which is a real-time thing. You've got lots of time to compress. When somebody wants that data back, you've got to decompress it really quickly. And that's where computes come in, that's where multi-core comes in, that's where as many cycles as you can give it. Yeah, you hate that. So you mentioned Memrister, so thinking about Flash, as enterprise people, we probably never would have selected that technology because all the gymnastics that you have to go through to make it not wear out and make it reliable. So do you feel like these other alternative technologies have potential given the volume that you get out of Flash with the consumer markets? Well, I think for the next generation, everybody who's anybody has got an ore and a water on that one, whether it's Samsung, Micron, Heineck's, HP. So I think we're going to see some competition and we're going to see the best technology emerge. And more importantly, since all of the big guys are involved, it will get to volume at some point because it'll get, part of the volume curve will get driven by the consumer device use of it, which is obviously a much smaller configuration, but you don't have to use as many chips as servers when you can drive the volume, that a tablet or a laptop or a phone can drive, and it'll then come up into the server world. And I think maybe be close to an order of magnitude decrease in the price. And the other one that's interesting that nobody ever thinks about, we all talk about real time and we all talk about these in-memory configurations, but NVRAM flash is persistent. And as people start thinking about what persistence means in the server world, it translates into reliability and availability. Because now it's in memory. When you go down in crash, you can reboot from an image much, much more quickly because you've literally got that in memory. You don't have to load it from a storage device. Plus is the possibility we get rid of what you call the horrible storage stack. Yes, yes, yes. You guys remember? Taurus, oh, that was beautiful. So I want to ask you about virtualization. Actually, we can talk about the storage stack. So several years ago, a lot of the Wall Street analysts that many observers saw the virtualization trend and you're running the Intel data center group. And they said, wow, Intel's in big trouble here. Because all that, they survive because it's all this underutilized capacity. Wow, they're going to be in big trouble when the world virtualizes. Exactly the opposite has happened. What did people miss? What's actually happening? And what's the future hold? Well, I think they missed a couple of things. And I'll try to remember to take them in order. The first one is high performance computing, which is growing really fast and becoming an increasingly bigger share of the market. And high performance computing, virtualization is like an enough amount of those guys because they don't want to give up one cycle. They want all the performance. So you first have one big chunk of a fast growing market that's not virtualizing at all. So what's 30% of the market these days? And so you've taken that off the table. The second one is that I think you've seen what I would call the straightforward virtualization happen. The well-managed, well-defined workloads like SAP, like infrastructure as a service, where you're provisioning an office or an exchange server or something like that. And I think what you haven't seen a lot of virtualization in is serious production database. Because that kind of like HPC wants security and wants, it hasn't bought in a multi-tenancy yet, can't really get enough really big virtualized memory areas because if you're putting a lot of the database in memory, you're lucky you can do it in the main memory you got, let alone have multiple VMs on there. So that's kind of another chunk of the market that hasn't really seen as much value from virtualization as people thought. So the more IO bound you are, the less you've tended to want to do it if you've got service level agreements. And as I said to somebody, if you're standing in front of an ATM, you don't want to be running on a VM. I don't know about you, but I know. What about the network and virtualization angle? Because OpenFlow has been out there, a lot of the big data folks are saying, it's not us, we've got cores, we've got cores on every piece of flash. We've got this thing re-architecting that's beautiful. We're going from five minutes to five seconds of app, but the network. Well, there's no issue with us. We had Broadcom said, no, there's no network problem. How do you come out of the elephant in the room? That's the one piece of technology I lay awake nights worrying about because I'm unfortunately old enough to remember that it was really DARPA and a bunch of government people we don't want to talk about that gave us the internet that we have today and paid for the whole backbone and the whole infrastructure. And it's not just getting fiber out to my house, it's who's taking that backbone to the next tier. And I don't know about you, but I live next door to Los Gatos. And when people are watching Netflix movies, you know the web is being consumed. Everything slows down. I talked to a partner of IBM just yesterday morning who does private cloud deployments and he's down in the Atlanta era. The sonnet rings are maxed out. I mean, we're talking like, this is not like this dullest dark fiber laying around. I mean, this is like this capacity problem. And they got now LTE 4G and 5G coming around the corner. It's going to suck more bandwidth backhaul. And we don't have a pricing infrastructure set up that's letting these companies invest in that. I mean, that's the fundamental limitation, I think, which is that until people can see how they're going to recover the cost of a Ryzen or an AT&T or anybody isn't going to jump in and take that infrastructure to the next level. You're absolutely right. You got Comcast who invested in that infrastructure. Again, you can go on either side of the neutrality argument, but Comcast just watches Microsoft Xbox but announced MLB, NBA, ESPN, live TV on the Xbox. Right. Very disruptive. We all know, I don't know about you, but we all know what happens to the internet during March Madness when they've actually literally had the cap, how many people can watch a game and you can't get on until somebody gets off. And I was saying to someone like, isn't that great? You can cut the cord. I'm like, but yeah, but the cord feeds the Xbox. So I go Comcast, they're using deep pack of inspection. They're going to throttle that thing down to the crawl to the crown. That gets me back to the numbers game, you know, which is unless they can charge more for it, they're going to throttle. Because the only way you can afford not to throttle is you got to raise your prices to get their return on your investment back. It's a real difficult thing. Cause now the lobbyists, we've lost this battle in the last mile with the, when we were at the SEALAC days, you know, they get choked, all the innovation got choked. So I'm worried about the network too. I'm worried about the network too. I'll tell you another funny network story. I was talking to a guy from a major research hospital. So you know the tier I'm talking about. They shipped data off to a national cancer database in the cloud with genotyping and tumor typing information. Do you know how they ship it? They looked at all the options using the network. They put it on hard drives and give it the FedEx twice a week. Shipping on the truck, it's faster. Yeah, exactly. The greatest bandwidth in the world is the Chevy truck. So setting aside the horrible storage stack, whoever thought hard drives were going to emerge is the interchange medium of choice. We still do sneaker net in our house. Our kids go to my laptop and put their, they email the file to themselves and they go to my Gmail and my thing and get a Gmail printed out there. It's like ridiculous. I wanted to ask you Pauline. So I was at a meeting at IBM, you know, Steve Mills. And so it was shortly after the Oracle did the acquisition of Sun and he basically said flat out, spark is dead. And of course we know the X and X, the Oracle stands for Xeon. We've joked about that several times in theCUBE. So it looks like games over, you know, x86. And then all of a sudden you see HP do moonshot. You saw Dell just announced, you know, low power. You guys have Adam. There's this new disruption coming. Talk about that a little bit. Well, first I would say there's always a new disruption coming. This is just the current disruption. So relaxed about that. No, I mean, I think we watch it, you know, and we are making some investments from everything from our entry level Xeon parts to, you know, what we're doing with Adam longterm. Cause we think it is a quotient of the power envelope and the density envelope and how many computes you need. And it's certainly true that, you know, there's a theory out there among some people that when you're looking at kind of generic web hosting and you want kind of the minimalist configuration that this is going to be the way to go. Now, we haven't exactly seen these things flying off the shelves. In fact, we've been trying to get a hold of a couple of systems to benchmark. And we haven't found anybody who'll sell us one yet. So I am a little dubious about this revolution that's happening. If we can't put some cold cash on the table and actually buy a box from anybody. But I don't think we're going to be surprised about it. We're going to learn about it. We're going to benchmark it. We're going to look at how it looks as a web server, how it looks, you know, for Hadoop, how it looks for other things. Because we think that there are places where people want a richer experience and are going to want to do more. I don't think, I think the IPDC guys have got these homogeneous, you know, hosting workloads. But how much of the real world that's out here has actually got a load, you know, a workload that's that homogeneous and that tightly characterized. Yeah, so it's maybe five to 10 years from the point at which we start to see that, right? The internet guys generally lead that. Right, right. And so is that a reasonable time frame to think about this, really having an impact? I think it's certainly in the five year space, also because it takes a while to develop the infrastructure. I mean, you know, it's one thing to put up an exchange server. It's another thing to run everything that you're used to running. And I think that's going to be a little bit, I mean, some of us who have had alternate architectures know what creating an ecosystem is like. I have one. Hold on, I have one. We're just going to ignore them. I have one final question, one final. I have a final question. Let's just keep going. So cloud storage. So I did a little unscientific poll amongst salespeople for a tier one manufacturer. I won't say their name, three letter acronym. And I said, how many people have customers that are actually using cloud storage? You know, mainstream customers, general purpose kind of customer, not the unique houses in Hollywood, you know, those kinds of things. No one raised their hand. So where are we with cloud storage? That's an obvious advantage. You're seeing, you know, the drop boxes in the box that put people in their cred down, putting the files up there. But actually- No hands went up? No hands went up. From a cloud enterprise standpoint, no hands went up. No surprise to me at all, because my old boss who's now doing PC's, Kirk Scalga, has been asking that question for the last six months. And he personalizes it. He says, am I going to put the only pictures of my kids in the cloud? Or am I going to use the cloud for backup and have, you know, two other copies of my kid pictures here in the house, you know? One on the USB drive that I can put in the safe, you know, and another one on the PC. And I think that's the crux of the issue. I think that for backup, it's probably going to emerge. You know, like my brother-in-law is a lawyer. He's a small business. He's always going to have what he needs in the four walls of his office, you know, a battery backed up and everything else. Because they're, you know, particularly for him in the legal business, there are confidentiality issues, there's security issues, you know, you might use it as a backup source. But I, you know, one of the interesting things is you got Microsoft coming out with this dual-node server, you know, it's redundant, it's backed up, it's mirrored. You know, that's the Microsoft, aside from Azure, that's the other answer to it, which is do you sell an SMB box that gives people more local control of their data? And that's the tug of war, I think. That's the tension of if you're a business and your ability to boot up tomorrow morning and serve as a customer who's standing in front of you is your responsibility. Where are you going to put the data? And I think that is the use case. It's backup and archive. It's great, you know? You have a lot of data that we want to just get off site, great, perfect for the cloud. Oh, good. All right, Pauline, we'll listen. Thank you very much for coming inside the Cube. Always a pleasure. Great to see you again. Great conversation. We'll certainly hook it with a beta account of the VDP Finder product stealth name. And great to have you on and we'll see you next time. We'll be right back with our next guest after this short break.