 Live from Las Vegas, Nevada, it's theCUBE at IBM Edge 2014. Brought to you by IBM. Hi everybody, we're back, this is Dave Vellante with Stu Miniman and we're joined by David Floyer who is Wikibon CTO, who's been hanging out in the analyst crowd or analyst meetings all day today getting a mega Kool-Aid injection from IBM of software defined and virtualization we're going to unpack it now. We really had a pleasure, David, having you on. I love these segments, thanks for coming on theCUBE. You're welcome. So, let's see, we were here in this very building a couple of weeks ago, we heard EMC announced Viper 2.0, Separating Control Plane and Data Plane. I had said many, many times that week I asked a lot of people, is it a way to consolidate the stove pipes or is it the future of software defined storage and the answer EMC gave was both. But in the near term, it's very clearly a way to consolidate the storage silos. Talk about what you heard from IBM, elastic storage, what they're announcing, is it different, how is it different? Both companies have the same objectives, same pressures on them, which is to work out a way where they can continue to sell what they have today and in the long run, get themselves positioned for a, quotes, true software-led storage or software-defined storage. So, from that point of view, we heard the same thing this time. We heard things shoring up the short term, some good announcements in the flash areas, some good announcements in the area of SVC. But the very interesting thing is that they put down a very strong marker for the long term. What we in Wikibon have termed server-san and what they're looking at is a global server-san, which, if they can pull it off, is going to be fantastic. But that's a long journey and a very ambitious journey and the thing that worries me more is how do they get volume in the short term to pay for that long journey? So, let's talk about server-san. So, server-san, you're talking about directly attached storage devices at the server level that are connected through some kind of high-speed interconnect and pooled through software to essentially create a global scale-out-san, correct? That's exactly right. Okay, and what's the benefit of that? You're eliminating a lot of complexity, presumably, but you're adding complexity too, aren't you? Well, the benefit is that the servers and the storage are closer together. So, for some workloads, you can get much lower latency and much further throughput. And the other clear workload where there is a short-term benefit is in areas like long-term retention or archiving where you want cheap and deep storage. And one of the very interesting things that IBM have put in this mix is a link to LTFS so that they're including tape and the ability to store on-disk or on-tape completely transparently as far as the user's concern. So, again, a very aggressive long-term vision, but there's going to need a lot of work to get there. Okay, now let's talk about the components of IBM's software-defined strategy. So, there's a variant of GPFS. Yep. There is a new name, Elastic Storage. Elastic Storage. There's some new technologies underneath. Well, there's Watson, right? They're bringing over sort of a mini Watson to help with the analytics and maybe data placement and what should go where. There's new technology coming out of Haifa. I don't know if you heard about this, the Storlitz. Yeah. Okay, so they talked about Storlitz, which is essentially a software construct, which is an easy way to customize the behavior of an object. Sounds like metadata. Metadata, yes. I mean, essentially GPFS is particularly good at managing the metadata. Okay, and I was saying off camera, also there's the technology from SVC, which you were saying, well, yes, but, but I mean, isn't SVC how IBM's going to virtualize the underlying infrastructure or did I get that wrong? Well, that's one option. The SVC is a great box at hiding, providing its own set of storage services and putting those on top of the arrays underneath it. But if they're going to be successful with the elastic storage or the service end, that has to have its own set of services. It can't rely on another set of services above it. That's, that's what I mean. Yeah, so let's talk about that a little bit. So EMC strategy, I'd like to start a harp on EMC, but they sort of laid it out. They said separate the data plane from the control plane, the control plane, or the data plane rather, is a bunch of EMC boxes. Right. And then they put Vplex in there. That's their sort of virtualization engine appliance. Not as mature as SVC, obviously. IBM's a little different. You don't need to buy an IBM box in order to get those data services, although do you? The Vplex does not have data services. Vplex doesn't, but it allows you to virtualize all the EMC boxes, right? VMAX or? Vplex is really active, active. That's what it does for a metropolitan and SVC can do the same job. Sure, right. And it's also a virtualization appliance, essentially. Yes, but it's not very mature. No, no, I'm not trying to critique it, but yeah, we can point that out. I mean, SVC is a decade old and Vplex is new, but my point is that with IBM, you don't necessarily have to buy the DS8000 or the V7000, whatever it is. But if you want the stack, you got to get the SVC. If you want the complete stack. So that's how companies, am I getting this right? This is how companies, IBM and EMC, particularly have put down products, so that's why we can at least evaluate them, how the legacy sort of storage industry is going to get from point A to point B. We're going to say we have stack, right? Our stack is, in the case of EMC, it's the data services behind a VMAX or a VMAX. In the case of IBM, it's got SVC. Right, and it'll take those services and it'll put them into, as a set of stacks, which is virtualized, it'll put that into the elastic storage. And then, meanwhile, they're working like crazy to build up their native stacks in the software defined stack. Just do what you're taking on this. Yes, so David, we've talked a little bit about this, and I was at OpenStack last week, and comparing contrast, EMC, they've got that viper. It's interesting, elastic storage, it's elastic cloud storage from EMC, so very similar names. Everybody's elastic or agile or adaptive, seems to be the buzzword of the storage industry lately. So EMC builds all of their services. They have the viper, and they've got their block, their file, their object, scale.io is delivering some of it. As opposed to IBM, OpenStack's going to provide some of the services in there, and they even rename one of their products to have OpenStack in it. So is it fair to say that EMC is building all of their services, and IBM is leveraging OpenStack a lot more? I'm sure there's a little more nuance than that, but dig into that. There are a lot of different ways they could go. So EMC are very clearly coming in at the Cinder level or at the Swift level, and they're allowing the arrays themselves to do the services. So their strategy is very clear that that's how they're doing it, at least for the moment. And then they've got Scale.io as a background, but all of those come in via viper to the Cinder level or the Swift level. Now, other sets of services, for example, in Solid Fire, come through to the horizon, where they're using the orchestration and the services from OpenStack itself. Now, IBM, I asked this question of IBM, and they say they're coming in at the moment at the Cinder same level as EMC. And clearly, they've got a lot of their own software, and they want to take advantage of their own software and put that as an alternative inside the elastic cloud. So in that regard, there are more similarities and differences in IBM and EMC strategy. However, my next question then is Open. Who's talking the Open game and who's actually laying down the resources? So if you've got a spectrum of total commitment, and on the other end is lip-marketing lip service. I mean, Dave, I can comment on that a little bit. So at EMC World, I asked David Goulden directly, what's your position on open source? And he said, fundamentally, open source is not part of our strategy. We want open interfaces. We do want to give customers choice and flexibility, but the federation is the model. Well, Jeremy Burton said the same thing. But the federation being the model, their open-source strategy is pivotal, okay? IBM's is different, I think. 380 contributors to OpenStack. So that's not lip service. No, not at all. HP as well, HP props to probably even more committers, I think, based on the numbers they're throwing around. I don't know, you never know. It's about 500, I think, for HP as well. But anyway, 380 at IBM, let's say 500. That's some good engineering resources. I didn't hear any EMC. No, no, I have never heard any numbers from EMC. I mean, I don't know if the number is zero. No, no, it's not enough to brag about. Yeah, I did get an answer from EMC on that. They have a little bit of work getting on. They answered where they're working into Cinder in a couple pieces, but no booth at OpenStack, very little. Right, so that says to me that it's marketing lip service that EMC is giving OpenStack for now. It's deliberately coming in. Because it's a big franchise, which is VMware. IBM doesn't have that to lose. They already lost it. HP already lost it, so now they get a counter attack. Okay, so it makes perfect sense. So these are interesting sort of machinations going around on the chessboard. But there's no doubt in my mind that IBM will make Elastic Storage much more open and be much more comfortable with going in eventually at the horizon level, for example, in OpenStack or the CloudStack or the number of other OpenStorage. So OpenStack is a big wild card for the incumbents. Generally, specifically EMC, I would think Cisco as well. I mean, not so much. I mean, is that NSX is the wild card there? This is one for Stu, actually. So, I mean, David, we interviewed Luke Tucker at OpenStack, sir. If you're asking how much does OpenSource play in? We always joke Cisco's about 100 companies. And there's a couple of companies inside Cisco that are heavily committed to OpenSource and driving it, but they're fighting against a lot of battles internally there. So today, if you were to ask me, which is more open VMware with what they're doing in NSX or Cisco, I'd actually have to give Cisco credit because they've got more people involved, had a sizable presence at OpenStack as opposed to VMware took NYSERA, which had OpenVSwitch and helped create the foundation for Neutron. And they've really pivoted to be more open interfaces as part of that Federation strategy. As you said, Dave, Pivotal is the open source piece of it and EMC and VMware are gonna more provide those interfaces. So didn't VMware donate NYSERA code to OpenStack? So it was the foundation of what's there, but if you look at contributors, there's a lot of companies that are contributing there. And actually at OpenStack, Neutron is probably the most broken of them and they're not blaming the VMware folks and the NYSERA folks on this, but it has been correct donated as the platform and everybody's working to fix it. I want to do a quick aside. David, I don't think you were there this morning when I asked, too, I think you were there. When I asked, somebody asked about OpenStack and I followed up with the Red Hat question. So it was interesting. We've been following this story. In case you haven't been following this story, Wall Street Journal broke a story last week that said basically that Red Hat was holding back support for its Red Hat Linux distribution on any non-Red Hat OpenStack distro. That's not true. Red Hat will support any of its Linux distros anywhere. What it won't do is certify them. It will only, so far, it's only certifying Red Hat's distribution, which for instance, Dell uses. Certification versus support. So it will support narrowly Red Hat Linux. It won't certify the whole OpenStack platform or product as we talked about a week about the other day. So I asked IBM, I said, what's the story? What's your take on this? And they said, well, you know, that's a question that you should ask Red Hat. So very clearly, IBM would like to see a collaboration with Red Hat where Red Hat certifies IBM's OpenStack platform. So there's clearly tension there. You heard Martin Fink or saw Martin Fink made a quote in the Wall Street Journal that he's probably rolling his eyes and regretting it, but he said something about Red Hat is bringing a new art form to closed open source. Red Hat shot back. Well, we'd love to see HP do some more in open source. So you're clearly seeing some tension amongst the committee. Red Hat is becoming a competitor of all these guys. Stu, anything you can add to that discussion? Yeah, so it's a bit complicated in sorting itself out, Dave, but you're right, from a certification standpoint, Red Hat is not open to everyone and is kind of using their 60 plus percent market share in Linux to kind of muscle in there. But the thing that I was a little concerned about is if you talk about putting Red Hat at the guest level on top of a hypervisor, they said, we certify VMware, we certify Hyper-V, and of course they certify their own KVM. So this is where if I choose kind of the majority of the environments, they are open, but if, for example, I'm a Rantis and I've got a lot of open stack environments and I'm not using Red Hat's KVM, they're saying, once again, Red Hat's trying to muscle their way into this room. Here's my take on this. And a lot of people will criticize them. Randy Bias was sort of, he was going back and forth to me on Twitter the other day. I was sort of neutral on this, but I sort of formed my opinion. Red Hat is now a company with nearly a $10 billion market cap. This is the big time, okay? This is about TAM expansion for Red Hat. They're not going to kowtow and say, oh, we're not going to get into your business because you helped us get to become a $10 billion, no. They're going to compete. You know, this is now game on, right? Open stack is a wide open opportunity for HP, for IBM. You know, maybe it's more of a threat for the VMwares of the world. And Red Hat wants a piece of the pie because it's got to continue to go. It's got to continue to deliver shareholder value. So it's going to be very interesting to see and why wouldn't you, if you're Red Hat, leverage your unique advantage. You know, Randy Bias can complain all day, oh, it's not in the open source spirit. It's all about the economics of the business. And Red Hat has a great track record on open source. I really don't think it's fair to criticize. I mean, it's fair. Everything's fair, but I'm not surprised in the least. Are you? No, not in the slightest, but going back to IBM. IBM have some real mojo in this area. They've been successful in the past that using open source, they have a good understanding of that business and they have to. I don't believe there's any way that they can maintain market share or grow market share unless they take that approach. So IBM's playbooks are starting to take shape. It's like pound a billion dollars, go hard on open source, bring in analytics, use some of our unique R&D, like Ride Watson's coattails, and then go hard after our install base and win that and suck in new customers as we can. I mean, look at IBM and the big data business. IBM's the leader. The leader in the business, yes, absolutely. And so, and they're applying that. And I mean, look at Tivoli. I mean, there's great success of Tivoli. So even though it's not a lot of, you know, attracting a lot of off-platform IBM, I think IBM's trying to use that same playbook in storage now, you're seeing it in flash. So I wanted to give you a last word on flash. Give us the bumper sticker on flash and Stu, you chime in. So IBM in flash, they've gone hard after PCIe. They're also going hard after Sandisk and Diablo with the DIMM form factor in that area. They've announced a low entry flash, which is nice. They're trying to get lower down. They've put a lot of compression into all of the flash. So they've got a good flash story. Where they're missing out is in having a flash only, a real flash only array, which can compete with the Pures and the other models. TMS is not that. No, TMS is a really high performance one. So it's great if you have very, very high performance oracle databases, that's what you want at that high end. But it doesn't go after that middle ground. It's not going after the fat middle, it's more expensive. It's a $3, they're up in the $6 and $10. So we get pure with the $3 billion market value. EMC is bundling extreme IO in and being very aggressive about that. You get the violin that's been out there. We all know the IPO story. HP is announced, but it's still trying to make inroads in the marketplace. Then you got the hybrid guys. Who's the leader? In terms of, at the moment, in terms of the fat $3, $3 a gigabyte pure is leading. So we're talking about all-flash arrays right now. All-flash arrays, yes. We're not including fusion and the other guys. Well, then fusion is at the PCIE. I have EMC that's shipped a hundred petabytes so far of flash. I don't know how that compares. EMC, I think it's 17. They've got a lot of server. They've got a lot of server stuff. Stu, I'll give you the last word. Yeah, and I have one more question for David after I finish, but on the flash piece, TMS might be hitting the high end, but from a revenue standpoint, EMC claimed that they had leadership with I think it was $67 million last year. And when I'm talking to a couple of IBM people behind the scenes, they said, well, I wish I could share what my division is, but I don't know that EMC is still the leader. So definitely trying to see the flash market sorts itself out between the hybrids and the all flash and everything else. So the last thing I had for you, David. It feels like the American League East. Yeah, yeah, yeah, yeah. It's great analogy, Dave. Well, when you're talking about SSDs, then clearly EMC with 30% of the market has a huge amount of storage. Yeah, yeah, sure, because they got the incumbent store. So they got the incumbent store. So David, the theme of the show is infrastructure matters. So we've been talking about modern applications and cloud and everything else. Does infrastructure matter? And if so, why? You've got to have an infrastructure which doesn't get in the way and can offer you whatever the application requires. At the end of the day, they said it, and I really believe it, workloads count, workloads of what matter. You've got to have an infrastructure which is fungible and flexible enough so that you can create the most cost-effective infrastructure for that particular workload. Okay, I think the equipment's overheating. The smoke is coming out of the boxes. We've been going all day here. All right, we're going to pick it up tomorrow. Jamie Thomas is going to be on. We've got more practitioners. A big focus of this year's Edge is customers. We love talking to the customers, getting their case studies. So David, thanks very much for coming on. Stu, thank you for all your co-hosting. John Furrier is flying in. He's probably landed by now. Let's see any minute. So we'll see him tonight and he'll be co-hosting with me tomorrow. So join us. We start off at 9.45 Pacific time tomorrow. And so that's it for tonight. Day one is a wrap. Thanks for watching everybody. Go to crowdchat.net slash IBM Edge. Check out the crowd chat. We've been running that all day and hit us up on Twitter. Thanks for watching. We'll see you tomorrow.