 Hi, everybody, we're back. This is Dave Vellante with Stu Miniman. We're here at Edge Live at the Mandalay Bay Hotel in Las Vegas. Kevin Powell is here. He's the Flash Systems Business Line Manager for IBM. We've been talking Flash a lot. Probably, I don't know, Stu, at least a third of the content, maybe more that we've touched on has related to Flash. We talk big data, we're talking Flash. We talk Jeff Jonas. We end up talking about Flash. In he Joe Suck comes on. We talk about Flash. Yeah, Flash is changing the architecture everywhere and touching everything. So Kevin, welcome to theCUBE. Really appreciate you coming on. Thank you, Dave. Welcome to have you. So Edge, the discussion last year actually wasn't much about Flash. You had some partners here. You guys were reselling some other people's products. That's cool. Bang, you make the TMS acquisition and all of a sudden, in typical IBM fashion, you're smoking. Got us off and running. That acquisition, kind of the first start tipping point, if you will, tip the iceberg in terms of getting the Flash ball rolling and really getting the value out to our customers. Yeah, and then April 11th, Steve Mills announces you're going to drop a billion into Flash. And he's done that before. He said that in analytics. He ended up putting in 10 billion. Right. He talked about that back in the Linux days. You know, I'm going to bet a billion dollars in open source. So you guys have a good track record there. And we've talked, we talked to Mike Hewn earlier today about, OK, what's happened since the 11th, but he underscored this big uptake, talked to Laura Gio, said, pipeline's good. You like that. So you guys had some announcements at this event. Talk about those a little bit and then we'll get into it. Sure. Really, the focus of this event has been extending the overall Flash portfolio. So on the April 11th event, we announced our IBM Flash system products, the standalone Flash arrays. But part of the focus of this event has been extending the fact that it's not just the IBM Flash system products, but the broader Flash portfolio across IBM. And so a couple of things announced in particular, some things around software. So some caching features, both in our DS8000 portfolio, we're actually taking the easy tier technology in the DS8000 that does tiering between disks and Flash today and extending that into the Flash in the server. So now we're connecting the Flash capability of the storage with the Flash capacity within the server and connecting that together with software. So that Flash capacity in the server is a PCIe connect? Correct. That could be an SSD drive form factor, 2.5 inch, 1.1 inch, or it could be in a Flash adapter form factor, so the PCIe cards. OK. And then so help me understand the connection between that layer and the DS8000 in this case. What are they doing? How are they talking to each other? And how does that translate into business value for customers? Sure. So first, it's kind of two layers, if you will. So first, there's caching. And then actually with the DS8000 we were achieving cooperative caching. So first, in the server at a caching, what the software will actually do is as you do reads and writes from the server to the storage, the server will actually determine, well, what am I most likely to read next and keep a small amount of that information local on the server within Flash within the server for a small amount of capacity, very high response time co-located right in that server. And so that's caching we actually could deliver. We did a statement of direction on software. We're going to deliver a system exit that will do caching, as well as caching now with the DS8000. So sorry, that's a read cache. That's a read cache. So it's not a persistent medium, or is it? It's a read cache, or a read cache right through. So any writes would happen, would write through to the persistent sand behind the data, behind the server. Which would go to the write cache in the array, and then eventually asynchronously trickle back to the spinning disk. Correct, so it could be to a storage array that is all Flash, or it could be to a storage array that was disk or even a mix of Flash and disk with a storage. So a storage array that's all Flash, obviously you don't have a destaging bottleneck. Correct. Storage array that's spinning disk, you're going to back up at some point. Right. And in the old days, you'd add more disk, maybe you're better off adding Flash. So we've been focusing a lot prior to this with the DS8000 on easy and store-wise portfolio as a sand volume controller, on easy tier, the ability for the system automatically to move data between a Flash sheet level tier and a disk tier for hot data. And what this announcement now enables is to extend that all the way to the server so the system can really prioritize what information is hottest, it shouldn't stay even on the server in Flash, what information is hot, but maybe need across multiple servers so it's on Flash within the storage, and then what information is colder, and I'll put on spinning media in the storage. Kevin, I'm wondering if you can talk a little bit kind of competitively. Last week, Dell talked about their Flash Cache technology doing something similar. A month ago, EMC talked about how they've got their server-based solution going down. We've dug through a lot of the tiering solution. IBM tends to focus on really trying to fully automate that technology versus trying to put too many knobs in there. How does this fit into that and what can you say about the competitive aspect against what you've seen out there? Sure, and that's where, there are a couple of vendors out there, they're starting caching is obviously not a new concept. So customers and vendors out there with caching, what's really key and different from our perspective is with the DS8000, it's cooperative caching. So what that really means is instead of just individual servers doing caching, the DS8000 and the easy tier functionality is aware of all the cache on the different servers and can act as a central intelligence to control all of that information. So if that centralized server needs to do a copy, a mirror, a backup, or it even knows one server has written the information and that's devalidated, it can tell another server, hey, your cache is no longer the latest, go update your cache. So it's taking the caching technology that you might deliver just on a server side and giving it kind of a centralized intelligence with the storage environment to coordinate multiple servers doing their caching as well as tiering within the storage itself. So do we get through all the announcements? Well, if we've gotten up through the announcements, if I look at kind of the Flash portfolios out there, there's the server-based solutions, there's that mixing between the server and the storage, you've got the Flash systems, which is kind of the point solution for all Flash. What about hybrid systems? It's been getting a lot of buzz, there's a lot of startups in that space, new architectures that has not just the one to two percent, but more Flash in there, kind of the Flash first when it comes to the rights. Does IBM have a play there? What do you think about that? Yeah, so in the hybrid environment, that's actually, we've been doing that in there, our store-wise in DS8000 since 2008, and the new Flash systems are kind of all Flash already. So to put into context a little bit, step back, really the IBM Flash portfolio, it's three different major categories. You have IBM Flash Storage, which is our IBM Flash system, or all Flash DS8000 or hybrid DS8000 store-wise products. You have IBM Flash software, where the capabilities we just talked about with things like caching, and then you have our direct-attach portfolio. So IBM Flash adapters, Flash drawers that attach directly to our power and our system X servers, as well as disk capacity. So the hybrid environment, we're still seeing a lot of customers' interests, where they're using the Flash, particularly with easy-tier technology of store-wise, Sandvine controller or DS8000, to optimize, okay, I have a Flash capacity for my house data, I'll see 200, 300% performance improvements over with my current capacity by putting in just five, 10% Flash capacity in front of my current disk capacity. Okay, so do we get through the announcements, or anything else that? That's the key, really, the software is a new piece, and then just accentuating the fact of the broader portfolio. So we talked a lot this week about customers putting a couple use cases. One is the application is doing the storage management, replication, data protection, whatever, and you drop a Flash system right in, and it's pretty clean. The other use case was dropping behind in SVC, we had one customer, Sprint was on today, we talked about that with Karima, Abdullah. Where do you see that going? Is that, are you guys, is, question, a lot of people ask, is that a stopgap? Are you going to try to build out your own stack with Flash systems, or is SVC the stack strategy? Can you talk about that a little bit? So there's two parts of that. There's the fact of current bundling buying the two products together and putting it, and then there's what's the right software feature stack, and the, while the bundle itself today is a stopgap that we're continuing to invest on, the core SVC code that we leverage across Sandvine Control Storewise is a strategic code base for us and feature set that we are looking to bring to the Flash system portfolio. So you'll see over time, and we're seeing the marketplace is the all Flash array marketplace really bifurcating the two aspects. One being customers focused on pure performance. They want the lowest latency, highest performance, they don't want any software features or anything getting in their way. They're purely performance driven. And then there's another set of customers we're finding that need high performance, need low latency, but they also need some of the enterprise feature stack, copying, mirroring, virtualization capability. And that's where today, we recommend the Sandvine Controller with their IBM Flash system products, and in the future, that's where we're going to be integrating those technologies together. So as a single IBM Flash system offering, you have the choice of pure performance with feature rich software stack. Okay, so when you say over time, you're going to integrate those, you mean as a, is that a packaging go-to-market strategy? Or are you talking about extracting the function out of SVC and super gluing it into? I'm talking literally about taking that software code base and bringing in, doing further performance optimizations on the hardware. So right now we have a system where you can use a Sandvine controller in front of a Flash system. We've tested them, we've tuned them, they work great together, but we also think there's places in the future where we can actually bring that software code indirectly and optimize it for that high-performance Flash hardware base and deliver even further integration over time. I see, okay, so make that code portable over time, and then of course, and then it fits into an SDS strategy, perhaps, it's behind an SVC, or like you say, direct connect to where the application does a pure performance and nothing get in the way, or something in the middle, which has a lot of legs as well. Right, so you have that choice and balance of absolute extreme performance, lowest latency, or I need high-performance low latency, but I need some mirroring, some copying, some other feature-rich capabilities. You guys bought TMS, you looked at everybody that was available at the time, you had your pick of the litter, so to speak. What were some of the decision points that went into that? First of all, how would you describe sort of the flavors, the choices that you had, and why TMS? Sure, yeah, actually I was involved at the Institute of Diligence, and that was a great experience. I think one thing we see in the marketplace, as in a typical environment that was up and coming, new ideas, a bunch of small companies coming out, maybe some VC funding, some ideas, but not necessarily a lot of companies that had really market experience, and that was one of the things we fell in love with Texas Memory System, they had over 30 years in the marketplace, they had a product that was outshipping, proven with customers today, and had that real-life experience versus just good ideas. So in some instances, we almost referred to Texas Memory System as a 30-year-old startup, because they really almost acted in that fashion, small team, very energetic, very quick from a development standpoint, but still real market experience, having been there, having sold in the marketplace, and dealt with DRAM based products originally, then flash technology, but solid-state storage devices really for their duration. So what do you think's going to happen in this business? You would expect a lot of consolidation, right? I mean, the market's really not going to support whatever. I don't know how many vendors are out there, it's got to be 50, 60, some way, shape, or form, and some sort of leading the pack, but you guys have made your move, EMC has made its move, see what HP does, Oracle has yet to make its move, although it's developing some stuff that appears internally. There's just, there aren't a lot of whales left to buy these companies. So what do you think's going to happen? I mean, do you think it's going to be like the tier 1.5 was, where they all disappear and get absorbed, or is this market big enough and hot enough to support some startups going IPO, or do you have an opinion on that? Yeah, I guess my own personal opinion. I think you'll definitely see the consolidation to particularly the major storage vendors, particularly in the enterprise marketplace. I think there's still places where we'll see innovation and software feature stack and capabilities that will cause new ideas to come up with new startups. So to me, it won't just consolidate, there'll always be fresh ideas, but it's becoming a scenario to us where flash is everywhere. I mean, you walk around an event like this and pretty much almost everyone has flash in their pocket with a cell phone or some other device. So it's really getting a point of a matter of not are you going to use flash, but how much are you going to use and are you going to use it in your server, your storage, where are you going to integrate it in? So we do see that consolidation in the marketplace. Yeah, and we were talking before about the DS8000 and basically how that's incorporating flash and you're seeing that across the industry, but part of me just says, well, at some point, why not just drop in and all flash array and I'll be happy, like my laptop doesn't have a spinning disk. That's where I'll never have a spinning disk in. I've sworn myself off spinning disk. Not that there's not a role for it. There certainly is, but the idea of short stroking spinning disks is sort of arcane. So am I overstating that? Is that like, you know, people have been trying to kill tape for 100 years, but... Well, and tapes are very good analogy. So just like tape, I don't think the spinning disk will go away completely, but as you mentioned, doing things like short stroking, 15K speed drives, trying to do the high performance disk segment as spinning disks, that is where flash is really taking over. And that's where we really see that market disappearing as a disk market and the high performance that moving to flash, but disk will remain as the larger capacity, small, you know, slower speed for the cold data. And just like tape's still around and still viable business for that cold as data, we really see that becoming the scenario with disk where you'll have a mechanical spinning disk for the large capacity and flash really being your performance layer. What are your, I wonder if you also have an opinion if you can talk about the whole notion of atomic rights. We've talked about a lot about that at Silicon Angle Wikibon. We get excited about it because of the potential to change the way in which applications are designed. I know you guys have, you know, partnerships with guys like Fusion IO and you've looked at that very carefully. What are your thoughts on that as far as the viability of that? Is it problematic and that essentially you got to change the way in which you write software? But or at the same time, it's appealing and alluring because you're bypassing the overhead of say the SCSI stack. I wonder if you could talk about that a little bit and share with us either your point of view or IBM's. Yeah, so I think that's one of the places where we are seeing innovation above and beyond just, you know, faster hardware, faster NAND capability with things like atomic rights. And I do think it's going to be disruptive to the software stack, but that flash overall is almost disruptive to the software stack. So we're already in a lot of conversation with our own DB2 teams and other ISVs and application vendors to say, okay, you've done all this work prior to now to avoid IO. You've tried to do in-memory database and do other things to avoid IO. Now you have high speed IO. Now you have all flash storage. How do you better optimize that? How do you start changing your application stack? So whether it's atomic rights or other innovations, I think there is going to be a change in the application layer and the OS layer to start to say, hey, we have to take advantage of this capability in servers and in storage because that IO bottleneck's not going to be. Yeah, and in-memory's cool, but it's expensive and it's not persistent. Correct. So those two things are problematic and presumably flash can solve that. But I wonder if we could have a follow-up question on that. So you've got a standard emerging NVMe, which looks like it's going to take forever to get out of the committees. It'd be great if we're here today and the industry could adopt it. So is IBM strategy to wait? Will you try to develop your own or no comment? Well, I can't comment specifically about that one, but I will comment. I mean, we're not waiting in overall, right? So we're investing a lot of money with the public $1 billion statement in the investment, but our R&D team, even prior to this, has always focused on how do we better leverage flash technology, how do we better leverage the application stack and we always as a company try to be a leader in terms of setting standards in the industry. Yeah, I mean, do you think, is it technically feasible? The problem with atomic rights is to get there, it's very disruptive. You can't get from point A to point B without essentially either ripping or replacing or completely changing the way you develop applications. Technically, do you think it's even feasible to have a technology that is not disruptive but still gets the benefits or even something that's close, maybe even some kind of hybrid? Is that even technically feasible? Well, I think we are seeing a move where there's a lot more focus of the application stack on fully integrated solutions and that's the easiest place for something like that to start. Not going out and trying to deliver a standalone storage device and you're not sure what the OS, the application environment, but things like peer systems where our peer data for operation on the lakes today uses a large flash capacity, places like that where we are delivering the application layer, the OS layer, all the way down to the storage, I think that's one of the places where you'll start to see that test out where you can deliver specific industry applications, leveraging that capability because you have the complete stack and I think we're uniquely positioned to be able to deliver that complete stack solution and then through our customers. All right, Kevin, I'm going to give you the last word here. You're talking to customers, they're sort of questioning, ah, flash is expensive, I'm not really sure where to use it or I can use it a little bits and pieces here, but I'm interested, I want to get started. What's the advice that you give to those clients? I think the biggest thing I would say is, you got to try it. So we have established flash centers of competencies around the world, benchmarking centers where customers can come in and get their hands on. I've been through multiple product generations across server and storage and I've never seen anything like this that has the success ratio. When customers actually get their hands on it, they run their production data on it. We heard from some of our customers earlier today saying, well, hey, I put it in our performance in our test environments, but then our production even performed better than I expected. And so that's where I say, whether you're starting small as a flash adapter or ready for an all flash storage array, testing out flash really in your production environment is when you can see it for real. Drop it in and hang on for the ride. Kevin Powell, thanks very much for coming to theCUBE. Really appreciate it. All right, keep it right there. Everybody will be back with our last guest and then we're going to wrap. Vincent Su is coming on. He's the CTO of the IBM Storage Business. Storage is changing. We've been covering it here for the last two days, IBM Edge. This is theCUBE. Keep it right there, but right back.