 Hi everybody, this is Dave Vellante of Wikibon.org and we're back with Pat Gelsinger. Pat, welcome again. Thanks for taking some time with us. Hey, it's always my pleasure, Dave. Thanks so much. Good to see you. We're here. We're going to talk about flash. Now we got together last year at EMC World. You were really gracious with your time. You probably spent close to an hour with John and myself. It was fun. That was good. And one of the areas we talked about was what was known at the time as project lightning. Yeah, I told you we were going to do it. Yeah, you showed a little leg at the time and gave us a glimpse. And now we're here and it's come to fruition. It's called VF Cash. But so before we get into that, maybe you can help our audience understand what's happening with flash in general. One of the things that we've written about a lot at Wikibon is how server function migrated out of the host for the last 20 years. Of course, EMC had a big role in making that happen. Sure. Because you watched it from the other perspective at Intel and now it seems like function is starting to move back. Is that an accurate assessment and what does that mean for the whole flash trend? Well, we do think that there's a couple of big trends. One is the whole idea of this virtualized data center where in fact you're creating more flexibility of what happens on the infrastructure where we can start moving stuff around. We also see that there's this huge IO gap that's emerged where, boy, if there was only a little bit faster CPU versus what you could get from a disk drive over the network, it didn't matter a lot. Now that gap has become enormous. So there's this huge tension that says drive down latency, give me better performance. And all of a sudden the question is can I use new technologies like flash in innovative ways to close that gap? And that's what VF Cash and Lightning is all about. Yeah. So the point being the disks actually are so slow relative to things like flash. Yeah. It's just incredible. Moore's law just keeps tripping along at this incredible pace so compute keeps accelerating. But while density has improved, guess what? Access times haven't changed hardly for a decade. So do you see the, it's always been cost per bit, cost per gigabyte is the primary metric that customers use to evaluate storage. Do you see that changing to the things like cost per IO? Well, always there has been a cost per IO question. You go to databases, people have always and whether we're short stroking drives or other things to give us more IOs per whatever metric. But there have always been the cost per gigabyte questions as well because the day to day lose continues to outpace Moore's law, continues to outpace aerial density improvements. So we know that cost per gigabyte matters, but we also know that cost per IO or raw latency matters as well. And it's really the question of how do you do both? And that's what EMC is all about. So we're going to get into VF Cash in a minute, but I see it as much software as it is hardware. Can you talk about that a little bit? Yeah. The value that EMC is bringing to this is software, services and support. Right? You know, well that's what we are bringing and we're combining that with a multi-vendor hardware environment where we'll pick the best in the industry. Our lead partner is Micron with VF Cash. But our value is very much the software layer that we're putting on top of best-in-class hardware underneath it. Yeah. Let's talk about that a little bit. I mean specifically we're talking about you managing the entire hierarchy from server flash all the way down to spinning disk. It seems like it's getting more granular. Talk a little bit about that extension of that hierarchy and EMC's role and vision around that. Yeah. That really is the heart of why our position here is differentiated. You know, if we back up a little bit when we first put flash into our storage arrays, everybody says, hey, that's pretty cool. But now I got a provision, yet another kind of disk drive essentially. And then we say, oh, let's do fast, this automated storage tiering. So we can move it from SATA, right, fiber up to SSD. And that was pretty cool. And the ability to move up and down that hierarchy, right, has been highly successful in the industry. And, you know, we did 25 petabytes last year, right, of flash shipments largely based on that technology. But now the question is, can you go all the way, all the way to the server? And that's exactly what VF Cash is about, is extending that fast hierarchy, not just in the storage array, but all the way to the server as well. And being able to have intelligence, right, from the server storage, right, all the way into the array storage, all the way into the most cost-effective storage available on the planet. So do you think all active data will eventually land on flash? All performant data will end on flash, right. Everything that needs performance goes to flash, and everything else will end up in the lowest cost tier possible. And it really does become this increasing bifurcation, right, of the storage model, right. You know, in the future, there won't be fiber channel, right, in terms of fiber channel drives. Because performant drives end on flash, you know, which means everything else is colder, right, which will end up on the most cost-effective SAS or SATA available. So let's talk about that a little bit, because you've got obviously a lot of data on Symetrix. It's the high performance, high availability system, you know, running the world's banks and insurance companies and all kinds of other, you know, retail operations, et cetera. What happens to that layer of the storage hierarchy? Does that move into flash, that function, or what kind of trends do you expect there? What we'd expect is, you know, if you looked at Symetrix past, it was almost all fiber channel drives, right. If you look at Symetrix forward, right, it will be a combination of SAS and SATA drives for lower cost, right, of solid state within the array, right, for performance within the array, and then intelligence to the server tier or to the VF cash tier. And that will all be managed by VMAX in the future, right, those three tiers of storage, the most cost-effective spinning media, right, you know, flash within the array and flash in the server, all managed through VMAX through a consistent set of fast policies. So EMC's vision in VMAX, and I presume VNX as well, because that single point of control that's actually managing the entire hierarchy, ensuring data availability, data consistency, and performance, and matching the right data with the device characteristic. Yeah, that's really the magic of VF cash, right. It's not just some other island, right, that's not managed, that's not coherent, not persistent, sitting somewhere else in the data center. It's just another, you know, piece of that whole storage hierarchy, all managed, all serviced, all being delivered through this consistent set of policies up and down the hierarchy. And from a user's perspective, it's all persistent, it's all HA, it's, you know, all of those characteristics that they know and love about our storage, they haven't given anything up. From the IT's perspective, oh, I know how to manage, I know how to service, I know how EMC supports this kind of stuff, you know, it is mission critical from his perspective, right. And from the end user's perspective, it's performance at lower cost, right. Everybody wins in this question. So let's talk about VF cash in a little bit more detail. So why VF cash and why now? Well, you know, we've been working on the idea for the last, you know, year and a half or so, this idea of how can we deliver, right, flash in new ways in the industry. Obviously, step one was just SSDs in our arrays. Second was fast, right, and delivering this automated tiering. But customers want more. This IO gap continues to expand, right. And this multi-core X86, I did a few of those in my career, right. You know, has just continued unabated, right, that IO gap has continued to increase. And so the ability to put a flash tier, right, a caching tier with flash into the server just gives another couple of orders of magnitude improvement in many use cases that just is so valuable to customers who are out for performance and out for reduction in latency that, you know, now's the time. So I wonder if we could talk a little bit about some of the choices that your customers make. EMC landed a haymaker was how we described it when inside of the array you included enterprise flash drives. Nobody really expected it. It was an amazing announcement because it really woke everybody up as to the, hey, flash is here and it's now. And EMC really shot ahead in the mind share of the audience. And it was, you know, really a great innovation. Over the next several years, we've seen other innovations from startups, maybe from some of the systems companies, people putting flash very close to the CPU. How do you differentiate from all those other activities going on out there? You know, we think of every other one of those as taking this new technology and working on innovation around it. And that's exciting, right? You know, it's just a thrilling period of the industry where a major new technology emerges and lots of innovation is enabled as a result of it. But when you go to an enterprise customer, they say, boy, do I want another island to manage? Do I want to only have this being used by a couple of apps? Because it's not persistent. Do I want it closer to the server? Or do I want it in the server? And all those forced trade-offs on the part of a CIO or IT, I can use it in some apps, I can't use it here. To us, VF Cash is saying, we're going to give you all of those value propositions without trade-offs. It will be EMC support. It will be enterprise grade. It will fit into the storage hierarchy. That's HA, that's business continuity, that's service and support. As you have it, no trade-offs with the maximum performance of any product in this class in the industry. This just wins because it's a better product. Yeah, so what do you see as Intel's role in all of this? I know they're a good partner. You obviously know the company very well. Where do they fit? Well, you know, I mean, Intel deserves enormous credit for the basic innovations that have occurred, even going all the way back to eProm and Flash. I mean, they have been a core technology innovator in this space for many, many years. Part of the magic of VF Cash is being able to take PCI Gen 3, the next generation of the IO bus that Intel has enabled, really being able to take advantage of that performance and latency inside of the server. And Intel's doing great work, not just with their own Flash products, but more importantly from our perspective in the platforms themselves. So, you know, they remain a key partner for us in many, many respects and we think they're going to be enabling more and more use cases for how Flash gets utilized in the server as we go forward. Do you think observers should expect that Intel is going to play a role in eliminating, you know, no offense to EMC but that horrible storage stack, you know, that's all built around, you know, spinning disk and then directly access the Flash as a memory extension. Do you see Intel as playing that role? Do you see EMC as playing that role? Or is it just too early to tell? Well, you know, we don't see first, right, and the very important point, right, because of the comments I made earlier, David, right, that, you know, the cost per bit of Flash never approaches the cost per bit in an aerial hard disk drive, right, kind of comparison. So, hard disk drives don't go away in our lifetime because we have way too much data to store, right, and the cost delta, right, between Flash and spinning media just never close, right. We don't see that for decades to come. So, you know, it's not a question of one versus the other, but how do you effectively do both, right, and when you go to the server side, right, it's all about access and performance, so Flash will be the dominant storage mechanism on the server side, and when you go on the array side, it's how do you take advantage of both, right, with hard disk drives always having the massive data on the array side just by the sheer volume and cost per bit comparison. So it's not either or, it's and. Okay, and with VF Cash now, you, however, are playing on the server side. Absolutely. My question is, are you now a systems company? Well, you know, we view it as extending the storage array, right, you know, we're extending our core capacity of being the storage and the, you know, at the other end of the wire and extending it into the server. I'm not in the server business, right, but I clearly am taking our value proposition much like we've done with HBAs for decades, or, you know, we've done with, you know, our different software technologies on the server side. Well, guess what, we now have a hardware product on the server side as well, right, called VF Cash. Well, now, of course, your colleague Paul Moritz is in the server business, software server, I guess. Yeah. So, what's the VMware play here? Talk about that a little bit. Well, you know, for VMware, right, one of the usage models that VF Cash supports is the virtualized environment, that's very important. And we're working with VMware to make Flash a better citizen in virtualized environments, to have better support inside of the virtualized, you know, the vSphere offerings, so that, you know, the Flash capabilities, which are, right, a different model of IO characteristics, can be better supported and utilized by, you know, ESX as well as by the hosts that they're running in those environments. Yeah, so now, of course, EMC had yet another, you know, tremendous quarter just announced this month throwing off tons of cash. All products really seem to be firing, you know, revenue growth higher than the industry average, which is great, congratulations on that. Thank you. You're painting a picture where, essentially, VF Cash and Flash is largely incremental to your existing franchise, if I'm understanding that. Can you elaborate on that a little bit and talk about when you really see this part of the portfolio really providing a meaningful contribution? Yeah, and we really are positioning of the product, how we're selling it to customers, and it's an extension of your arrays, right? And yeah, you can go use it without having our arrays on the other end, but boy, you know, from our sales guys' perspective, our customers' perspective, it makes VMAX and VNX just better, right? We're going to use those same management, the same support, the call home mechanisms, right, are all an extension of what they're already familiar with, and that really gives us a great advantage. We're also going to extend the fast APIs, right, and the hinting mechanisms through the server layer into the storage array, so the product is better as well when it's connected, right, with our storage arrays. So it really is an extension, make them better, right, and we're finding customers already saying, boy, you know, if I have that, you know, a combination of VMAX with this or VNX with VF cache, wow, that's a killer solution. In fact, I'm not going to consider some of those alternative architectures, right, like Exadata or something else. They're going to say, you know, boy, that combination is really fabulous to go with it. We do expect, right, you know, that we'll be a meaningful product line for us, right, but, you know, clearly, you know, this is day one. I don't want to get too far ahead of myself of the next blockbuster product. Okay, and you're talking about a very strong relationship between the storage array and the flash hierarchy. My question is around, and you know, me, Pat, I mean, I just, I'm firing on all kinds of different synapses when I hear you speak, and I'm interested in how you organized this project because you came to market pretty fast. You said you started relatively recently for an innovation like this, yet it has a Skunkworks feel to it. Was it a Skunkworks? And how did you manage the human capital of the innovation? Yeah, and we very much, to a great degree, you know, this was an internal startup, right, and, right, you know, the manager, right, of the team, I remember one time, you know, I said you are a startup, right, think about this like a venture team, you want the best architects, you want the best people, you want, you know, bodaciously aggressive goals for the team and, you know, just drive them just like a startup would, right, internally. And I remember he was, you know, came into a meeting one day, right, and, you know, I just ripped him apart. I says why are you in this meeting? Right, I says this is, you know, your standard big company kind of stuff. Why are you here? Right, you know, that's not how a startup works, right? You don't get distracted with this rest of the corporate stuff, right, that's going on. Get out of here and get back to driving your team hard. And, you know, it was a little event, but it just helped to reinforce the message, right? You know, we have to be nimble, we have to do internal organic innovation as well as the great track record and EMC has had in terms of acquisitions as well. So this has been thrilling that way. And now that it's launched, will you actually, you know, nurture it as you might? Will some of those startup characteristics continue or do you sort of just bundle it back into the business? Oh, no. Yeah, this is at least a couple of years yet that this is going to feel like an internal startup because, you know, we got a Gen 2. Just like any startup, man, we have a Gen 2 roadmap and a Gen 3 roadmap. You know, we are just driving that team for the second and third versions, right? We're also going to talk about a new project that's coming out of this team as well as part of the launch, right? You know, what follows lightning? Thunder, right? So we have a Thunder project as well that we're going to disclose, right, as part of the announcement. So, you know, another innovation coming out of this cool little startup team. So we're driving them hard. Fantastic. You want to talk a little bit more about Thunder? I mean, what can we expect there? What can you tell us about Thunder? Well, the idea of Thunder is that, you know, imagine I have, you know, five servers, right, each with a lightning card in them, right? Some customers might say, boy, can't I share my flash across the five servers, right? And instead of having it in each of the servers, just put it right next to the servers and make it like a little appliance. Now, the reasons you would do that is you'd be able to share, right, the flash footprint across multiple servers and some things like Oracle Rack, right, really fit into that model, right? When you do that, you could be persisting both, or you could be caching both reads and writes, as well as many customers have put in place high-density blade environments that don't have slots to put in a lot of flash directly into the blade server, right? So they say, boy, I like the characteristics of lightning, right, but I like it, right, so that I could utilize it with the installed footprints I have of blade servers. And in many cases, they don't want to go disrupt each server. They say, boy, you know, it's just easier for me to hook up right into the appliance right next to my servers. So that's what Thunder is, right? It's essentially an aggregation of lightnings into an appliance that, right, you know, will give a different value proposition, right, but building on all the lightning technology we've created. Yeah, so cache coherency across multiple servers, the ability to share, and presumably to support a whole new type of application. Yeah, and, you know, we see it in existing applications. They're going to be able to take advantage of it, but, you know, we think that people will see such enormous IOP capabilities coming out of this. They're going to say, oh, I could think about my application hierarchy different. Fun time to be a developer. Well, Pat Gelsinger, thanks very much for taking some time with us. Really always appreciate your perspectives. Thank you. It's a pleasure seeing you. Very good. Thanks for watching, everybody.