 Hi, everybody, we're back. This is Dave Vellante. We're here at day two at Edge. This is theCUBE, where we extract the signal from the noise. We go out to these events. We find the best guests that are attending these events. We bring them to you and extract the best information we can, package it up for you, our audience. You can tweet me. I'm at Dave Vellante. If you have specific questions, please fire away. So this is day two of Edge. We've been going wall-to-wall coverage. We're going to talk a little bit about Flash systems. Mike Kuhn is here. He's the vice president and business line executive of IBM Flash systems, and we're also joined by Jan Janik, who's also a vice president within IBM Flash systems. Former CTO. Jan, I believe you did the integration of TMS. Is that correct? Yeah, good. So gentlemen, welcome. Thank you. Good to be here. So, Mike, you and I spent a fair amount of time at the April 11th announcement. Had a very nice dinner with some of your customers and unpacked a lot of the information. So you guys made a bold move, got a lot of attention in the marketplace. Mills announced, hey, we're investing a billion dollars in Flash. When Steve Mills does that, you take notice because he's done that a couple of other times. So tell me, what's happened since April 11th? Oh, a lot. A lot of traction building around Flash systems in the marketplace. We had a lot of discussion, April 11th, as you remember, in New York City about the strategy and what we're doing there. So over the course of that time, we've built on a lot of client deployments with large clients. Many of them that are here at Edge just talking about their use cases and their deployments and the value and benefit of Flash. Yeah, so what's the shape of that curve? It's not flat, I take it. Much of the industry is flat, but I presume Flash is kind of up and to the right, is that? It's going to take off. Quite frankly, every client right now is looking at their storage infrastructure, they're looking at their application infrastructure from an end basis, and they're saying, where and how do I deploy Flash? So still a lot of people in the proof of concept phase right now, looking at major, major plays, but we're starting to see a lot of traction in this area. It's definitely taken off. So, Jim, when you take us back to the acquisition of TMS, what appealed to you about Texas memory systems at the time and how did it fit into the IBM profile of acquisitions? You guys are obviously pretty astute acquirer of companies. Why was it such a good fit? Really, two reasons. One is the technology that TMS had in Flash was the best we could find. A very, very deep focus on performance, latency, and getting the most out of Flash. The second piece, and so they had great technology and a great product, they didn't have a large sales force. So take that technology, add it to the IBM sales force, and it was kind of acquisition that makes sense for us. Yeah, you guys love those kind of acquisitions. Company with not a ton of sales and marketing prowess, and you come in, you do your thing, and then boom. Now, let's talk a little bit about how customers are deploying it. Mike, we heard, just heard from Karim Adela at Sprint, that he's basically putting the TMS behind, not an SVC. Maybe it is an SVC, a V7000, to me they're sort of the same basic virtualization technology as SVC, yeah. And we talked about, we had Eric Eiberg on, and we were talking about the 100 microsecond overhead, and Karim did a great job of explaining why the benefits outweigh that limited performance hit. So is that the common deployment model that you guys see going forward? Are you trying to build out your own flash stack? Or are you trying to maybe target situations where the application is doing some of that storage management, all of the above? Talk about that a little bit. So it really is targeting the applications, and it's really an application accelerator in terms of the specific use cases. Sprint kind of falls into the category around half of our clients fall into, which is they implement it behind SVC because they can drop it right into the infrastructure, they can migrate over to this new technology without having to rewrite the applications or change any of the operating procedures around cooperating copy services and compression and things of that sort. Other clients about the other half are implementing it as just tier zero storage, and it's really an application accelerator. So we go in there and we spend some time working with them about where they have latency in their applications, and really it drives immediate business value to their business outcomes in terms of end users or people within their application owners. Jay, can you talk a little bit about sort of the way we're going to see flash? You obviously see different deployment models. You see the PCIe at the server level, you see hybrids, you see all flash. You guys have an all flash with either an SVC or a direct install. You're seeing flash in arrays. You guys put flash inside of DS8000s and other arrays. How do you see this all shaking out? Are we just going to see this flash hierarchy emerge? Are they going to collapse? What's your angle on all this? Excellent question. So our focus is on that device to have the best optimized flash array for that set of end users. We will then take that device and integrate it either behind SVC or into our DS8K storage, XIV, and take that flash capability, put it on top of the capability of those systems that we already have. PCI cards is an interesting piece of the industry. There's a move to put PCI interface on two and a half inch drives. There will, I believe, always be a place for PCI cards. And I think there's an emerging space which we call in-server flash optimized storage. And that is very much a PCI card play in-servers. In-server flash optimized storage. So in my head, I always use the line, the best I always know I owe. And IBM obviously has a lot of server expertise. How are you guys working together with your server brethren to affect that, or are you? Are you guys sort of just focused on crushing it in the marketplace? No, we're working with the server team as well. As a matter of fact, here at Edge, we have Andy Monch on the Pure Systems team here as well. You're going to start to see our flash array, our design point for that whole flash array be part and integrated into the PureFlex system moving forward. So that's part of the future roadmap that we're starting to share with Fines now. Get some feedback on the future direction that we have going in that area. And we're getting a lot of interest, a lot of excitement. So how does that work? You're basically a division within IBM out of division, but an entity within IBM division, so you've got targets to hit as well. But you're also feeding other parts of the company with flashes. Yeah, we're a new business unit with inside IBM system technology group, which is our hardware division, as you know. Sit currently under the storage part of the STG area. But certainly, we do a lot of collaboration with the entire STG team, as well as the software group team. So you'll see us working with Natiza and Tivoli and DV2 as well. So how does the whole software defined thing affect you guys, Jan? Maybe you could talk about that a little bit. Where do you fit in that whole mean? Well, we play very well in software defined. It plays both in the in-server flash play as well as the external SAN attached shared storage. We will, of course, play well with open stack and interface into open stack to allow other users to add their capabilities and their IP on top of our flash. Yeah, you guys make a lot of contributions to open stack. I was just on the GitHub the other day, checking it out. IBM was right up there, HP, obviously, Red Hat, and some others. Solid fire pops up, a little solid fire doing some good work there. So it's interesting that you guys emphasize open source in a way that you do. Why is that so important? And why does that matter to customers? It's the future of the industry is where everybody is going. It's what they want to be locked into a proprietary system. They want to have an open and collaborative architecture. They want to be able to have been with the innovation of the whole industry. No one vendor can bring all the innovation into a client's enterprise. So open and collaborative is where we have been in the past and it's where we're going in the future. So, Jan, I wonder if you could talk in vague terms, because I know you can't give specifics. Just about the future's direction, where you guys are headed with this technology. So we're on a pace to deliver a new product every year. That's about as fast as the flash technology is moving. We need to keep up with that. We're on a track to double capacity, double performance, improve latencies, and add what we're calling enterprise class capabilities to the TMS storage. So you're talking about software capabilities. Is that right or not necessarily? Yes, software capabilities. Okay, so why is that important? I mean, why can't I just stick it behind an SVC and use the SVC stack or just drop it into an Oracle environment and let Oracle manage the storage management? What opportunities or markets are you able to tap by developing your own software? Maybe you talk about that a little bit. Yeah, not everyone is an SVC customer, but everyone needs to be able to concurrently maintain their storage devices or update code without bringing those down. You can certainly do that behind an SVC, but there are a set of customers who are not using SVC and they need those same sets of capabilities. So when you think about this notion of software defined, is SVC not necessarily a prerequisite? Well, there are a couple of vectors there potentially where I'll be able to actually access storage services. Sounds like directly from whether it's a flash system or maybe some other IBM storage system. Is that the right way to think about it? That is the right way to think about it. SVC will give you a set of storage capabilities that you may not get as software defined storage is emerging. Metro replication, for example, is probably not something you'd get in a software defined storage solution any time in the near future, but if you need that capability, you can put it behind SVC and get that kind of capability. Dave, in the same way that you saw, SVC sort of integrate within store-wise. We talked about store-wise earlier. You're going to see SVC capabilities integrating to flash systems as well in the future. Yeah, so I can infer from that things like compression which is going to help with costs, obviously. Copy services. Copy services, yeah, great. All sort of space-efficient capabilities that start to make flash increasingly attractive. Ambuj Galliologists, I think it was Ambuj on the cube yesterday, would say, if you're running 15K RPM disks, those days are numbered and we've been sort of calling that for a couple of years now. So you're starting to see flash just eat into, it's doing the Pac-Man into the traditional storage world, isn't it? Okay, so how does that all shake out? Are we going to see two tiers of storage, like all active data in flash and then they get a one-way trip to the BitBucket or are there going to be multiple tiers? Do you guys have opinions on that? Yeah, so in the conference, there's a lot of enthusiasm around flash, obviously. But I don't think anybody from IBM has the point of view that flash is going to kill off disk. This is not going to go away anytime soon. We guys still sell tape. We still sell tape as well. People call for that demise years and years ago, right? But if you take a look at the high-performance tier of storage, which is about 20, 25% of clients' enterprise data centers today, you're going to see flash start to significantly eat into that part of the marketplace right now. So they will still coexist with high-capacity disk and you're going to start to see a lot of advanced tiering that we have today in terms of moving the hot data onto the fastest tier of storage start to exist. Our clients are calling that tier zero storage right now. So you'll see multiple tiers of storage. That's the most cost-efficient way to go and flash is definitely going to take over in that tier zero space. So what are customer discussions going? How are they going? Because I mean, like I said, we had Karim Abdullah on and he's pretty advanced in terms of his thinking. A lot of customers either of just putting their toe in the water or haven't even brought flash in yet. They have the perception that it's too expensive. You guys are all saying, well, no. Actually, it's been an economic tipping point. Take us through a typical customer discussion. Well, actually the conversation has changed quite a bit in the last several years. A couple of years ago, we were having discussions about why flash. And now we're having conversations about where flash should play. So it's changed dramatically. So they're all thinking about this as part of their strategy. Certainly they have the perception of the acquisition cost of flash is about two X of high performance disk. So that's the starting point for many in this conversation. But it quickly evolves around, what are you trying to do with flash? What's the deployment case? And what's the end-to-end total cost of ownership? And that's where I think we work with you and other analysts and a couple really important pieces of information that we share with clients around. Here's the total end-to-end cost of computing with flash and it's driving about a 30, 35% cost improvement. Yeah, it's pretty substantial actually. I know the work that we did really related to the extent that you can balance out your system by using flash. Maybe it's going to cost a little bit more on the storage side, but you're going to save dramatically on cores. And it's going to have a direct hit on software license costs and as well maintenance costs. And I was impressed that Steve Mills stood up, and April 11 said, by the way, it's not just Oracle, it's DB2. And if we save you guys money because we're running more efficiently, that's a good thing. Exactly. He has to do more business with us, you know, that's. And you brought up Krim from Sprint a few times. He actually talked about the fact that it drives a substantial reduction in energy costs. Talked about as much as 85, 90%. And I'm not sure if you mentioned another cube or not when he was talking with you that Sprint was recently recognized as one of the best green companies, green data centers in the industry right now, I think the top and the telcospace. So they are certainly innovators, they're certainly on the leading edge and they're certainly making a lot of progress in that area. Yeah, he didn't mention that, but I'll remind it of Ed Walsh's keynote. I don't know if you guys saw that. So we have a room of 5,000 people said, said everybody stand up. He goes, okay, you're all a disk drive or a disk array. You're all spinning and you're throwing off heat and consuming power, et cetera. And then he had virtually everybody sit down except the guys in the front section. He said, okay, that's what happens when you drop in flash. Yeah, I was in the front right section. I was the flash guy standing there. Yeah, of course, you better. That's pretty dramatic. So that was kind of a good thing and I don't know how, but probably order of magnitude, it's pretty close. Well, that was a real life demonstration that we had done for a client in one of our flash centers of competency where we had said in a certain application to drive 45,000 credit card transactions per second, you could go from 5,000 spinning disks to like four or five flash systems. So it was order of magnitude improvement. All right, Jean, I got two last questions. So, Jean, I want to ask you, you're talking to customers, they're trying to figure out where flash, not why flash anymore, maybe where and how. What's your advice to those guys? The advice is go at this with a full, looking at your entire storage infrastructure. Don't go and just put in an 810 or an 820, which you can do and get some fairly dramatic results quickly. This is a journey to migrate your storage architecture from what it's been for a long time to what flash can enable you. Okay, and Mike Kuhn, you said before, you don't see flash under the illusion that it's going to replace wholesale disk drives, but you like to chest your cap because you're in a good spot with inside IBM right now. So what's your prediction as to how this whole industry shakes out flash's impact? What's it going to look like three to five years down the road? What's the impact that flash is going to have? So, right now, just a couple of statistics, flash systems, what we have with IBM flash system, as well as flash SSDs and disk systems, make up about 10% of the high performance disk tier right now in enterprises. I think in the next three to five years, you're going to see that making up the vast majority, over 50%, and over time in the future, I think the high performance tier is going to be all flash. Now, those metrics, are they revenue-based or terabyte-based or? Terabyte-based. Okay, yeah, and now, I guess, revenue would even be more substantial because on a cost per terabyte, of course, it's higher. Yeah, I mean, we're not going to have a faster 15K RPM drive anytime soon ever, right? So, all the major invention coming out right now with disk systems is about high capacity. So, that'll still exist, it'll co-exist as the tier two, tier three layer, but for the high performance tier, you're going to see that move quickly to, I think, an all-flash environment over the next three to five years. Yes, you must be excited, like I said, you're in the right place in the industry right now. What's next for you guys? You're on to more, you're visiting customers, you've got more trade shows, where are you going to be next? We've been, I haven't been, I haven't been home for one week and about the last, since April 11th, since we met about two months ago. So, we were over in Europe, we were over in Asia, we're talking to clients. Everybody is interested in flash, there's a lot of enthusiasm, obviously, behind it, but a lot of real hard work has to go into helping clients deploy. So, you're going to go home for a week, is that right? Yeah, sometime soon. Maybe July, August? Maybe this week, yeah. Very good, excellent. All right, Mike and Jean, thanks very much for coming on theCUBE. Really appreciate your perspectives. Good luck with the continued success of Flash Systems. We'll be watching. Keep it right there, everybody. This is theCUBE, we're right back with our next guest.