 Live from the Sands Convention Center, Las Vegas, Nevada, extracting the signal from the noise. It's theCUBE, covering HP Discover 2015, brought to you by HP, and now your hosts, Dave Vellante and Jeff Frick. Welcome back to HP Discover everybody, Dave Vellante and Jeff Frick. This is theCUBE, this is day three of HP Discover. Check out HP Discover.Social, it's our new social experience, social digital experience that we're overlaying on theCUBE at HP Discover. Milan Shetty is here as the CTO of HP Storage. Cube, Milan, Milan, good to see you again. Oh, great to be here, it's very exciting. So I got to say, congratulations on what's happening in HP Storage, really impressive what you guys have done. Thank you. You got to be pleased. Thank you, very much pleased. Very much pleased and more happy than I can be and thanks for the compliments. Coming from you, Dave, as a storage industry veteran and industry veteran, feels good. So here's the thing. I like to compliment folks, but at the same time I like to tell our audience sort of what I think. So when you guys said, we're not going to buy an all flash array company, we're not going to do what EMC did, we're not going to do what IBM did, et cetera. We're going to build on three par. I said at the time, well, HP's got no choice. Meg said, we can't make acquisitions, we got to pay down our debt. So they got no choice, but it's kind of a bolt on. You guys, well three par, I used to talk about bolting on thin provisioning. And you know what, they were right. And so the whole industry was really skeptical. Myself included that you could pull this off, but not only did you pull it off, it looks to be, if not the best, one of the best all flash arrays out there. I'm sure you think it's the best. So what did you have to do, CTO of that group? Take us back to the decision point when you said, okay, we've got to do this. Why three par, how three par? Yeah, so, as you said, I've been here before at the club, and I've been at HP now five years. And I joined HP about 10 months prior to what became the three par acquisition. And there were lots of reasons why we acquired three par. And one of the things which we did was we looked at the three par core operating system and everything. In fact, when we acquired three par, three par didn't have a very good flash story. They had a very solid modern operating system. They had a very solid architecture on the software side and also their ASIC side and everything. And they had done a very nice job of what I would call separation of concerns in the architecture, so that they could handle newer media, newer protocols. So now we added file also on three par. They had done this nice, very nice separation of concern in their core software architecture. And we liked that. We liked that when the now famous bidding war was going on between HP and Dell, we were willing, we were not going to lose that bidding war. Not because what they had, where they're shipping with their hybrid arrays or anything, but we knew where the market was going to go. The market was going to go with flash. The market was going to go with unified protocols and everything and that, the operating system, the three par operating system, we were willing to bet on, which is why we paid what we did. And in fact, we were willing to pay even more if the bidding war continued because we liked that operating system. Separation of concerns were fantastic in that architecture and having done storage and also the servers and also operating system a while. We just loved it from a technology standpoint and everything. So in the last five years, right after we acquired three par into 2010, we started working on making it flash capable. Nobody gave us a chance and it was kind of the love to play that underdog. We knew what cards we had. We had the secret play on, here's how we're going to change the operating environment. Here's how we're going to do it. And we released, it's a core three par architecture. Why it's good for flash. It's not only good for flash today with the dollars per gigabyte play which we're running, but as the newer technologies of flash and non-volatile memory technologies are coming, three par is ready for it right now. So the amazing thing about that is when you think about the three par architecture and the chunklets and the spreading, it's optimized for spinning disk. It's solving spinning disk problems, but what was it about the architecture that led you to believe that it was so appropriate for flash? I wonder if you could add some color to that. Yeah, sure. So the three par core architectures around the chunklets and everything was not only good for the disks but flash and also NVM. And I think the way the operating system was structured was a very strong storage operating system. I'll give three big architecture components to it. When they designed, they didn't just take any stack. They were optimization and efficiency was built into the operating system. Every single IO, they did a lot of performance analysis and everything to make sure that there was no overhead on any elements of the stack. It was a very thin operating system stack and everything, IO has to come in and go out very fast. It has to be reliable. So those principles were built in pretty nicely. The second is that in the flash world, it's a latency game. How do you get, it's like the packet delivery. You got to get from point A to point B as quickly as possible with the less route in the way. And the way they designed that platform was on platform is, is that they had algorithms built in the operating system to go from point A to point B, which in today's world would be called machine learning and the big data and everything. That was all in that operating system. It was all about getting to point A to point B as quickly as possible, as reliably as possible, was inherent in their core architecture. So all of the concepts around the machine learning and everything which the big data guys talk about. It's in the OS. So a lot of anticipatory intelligence. It was all in there, it was all in there. It was all in there, which is why we liked it from a media standpoint and flash, NVM and everything. This is the storage operating system of future. That's why we're willing to work with it. Okay, so it was the thin OS, it was the latency algorithms and that other sort of big data. That's right. When we didn't know it was big data. Exactly. And now HP so acquires 3-par. Now I used to talk to David Scott about this all the time. 3-par was barely a profitable company. It would lose a little bit of money and make a little money. It had to pour money into expanding its distribution channels to compete with guys like EMC. So it could never make any money now. And so as a result, it had to manage R&D very tightly. So it comes into HP. So you now inject sort of HP's resources in here. So what did HP classic bring to the table and what was the 3-par piece? Was that a collaboration? Or was it more, I hate 3-par guys, go off and do this. No, I think this is the HP brought the channel, the sales channel, the engineering expertise and everything people who knew operating system. So I'll give an example. So I've been here for HP five years and I can tell you about the customers. When I talk to customers, right? My year number one was, if I could describe in one word was my customer interaction was despair. Customers were do something in storage. Help. You used to be, yeah help. You were great. We still have EVAs, EVAs were great. They were the number one mid-range array until in the 90s, until EMC unseated them, get back on the horse, go on the race. So the year number one, I would describe at Discover was year of despair. And they were like, we want you guys to win. We're buying servers from you. You're number one in servers. But you know what you attach on the storage side is like back, so that was my year of despair. Year number two was year of hope. Well, and I call it year of hope because HP had acquired three-par and everyone in the technology community thought that that was the coolest operating system standpoint. And you know what HP did during that time was HP's got an operating system DNA. When we increase the three-par R&D staffing, we didn't have to train a lot of people how an operating system works or they could just learn the three-par operating system and say that this is a kick-ass operating system. We could just hire people quickly because they already existed at HP. So from an R&D standpoint, we had a big leg up. We just increased the R&D from three-par just dramatically in the first two, three, four years. People, just mostly people, mostly people. And the second element of that was they, we invested a lot in the equipment for testing, interop because storage business is all about interop and everything. So we put in a lot of money around the interop which three-par as a startup could not afford. So we added more people in, we added the R&D and most important, the most important aspect of all of this was customers. Customers wanted HP to win. Customers wanted HP storage to win. Customers wanted HP to regain the lead. The channel wanted to regain to win and that's like a perfect storm. Good technology, people. So you had despair, hope, what was year three? Yeah, so year three was optimism because what was happening was we introduced the mid-range. So all of the tier one features of three-par at the mid-range cost point. And that was the year of optimism is because there were customers who would walk in and go, don't screw up this acquisition. Good technology. Because we have some little bit of a checkered relationship with acquisitions in 75 years, some were good and some were not so good. So the hope was turning into optimism but there was a between year two and year three. There was the spirit of that just don't screw this acquisition up, we're interested to buy it. So year number three was the year of optimism. We bought the high-end features to the mid-range. So they were like, oh yeah, yeah, you're doing something good here. I'm optimistic about HP's future. So that was kind of a big trend which was going. Then the year number four was excitement because we moved from optimism that we brought the cost vector to the three-par but we introduced flash. And in the year number four, the excitement was that we made two successful transitions. One was the cost vector of high-end features to the low-end and then we did the disc to flash and all-flash array. That was exciting. Year number four was year of excitement. Last year I remember being here all-flash array three-par and they were like, oh, we didn't think you were, we were last year. And year number three, we were worried that you would screw up the acquisition in year number four. We were worried that you're not going to get the all-flash array but you were there. So that was the year number four. And year number five, which is now, is gratitude and year of high-fives. Customers are just going from despair to hope, to optimism, to excitement. To now it's the year of gratitude and they're saying, thank you. You have arrived. So a couple of things that, great. Thank you for that description. It was really useful and helpful. Couple of things that are, I think, conventional wisdom on HP which are, like you said, it's mixed, right? Acquisitions, some work out, some don't. The three-bar acquisition has been a home run. So HP actually can do good acquisitions and has proven that. The second is, not so much for HP but large companies can't do organic development. This, what I wanted to say, I told you I was off camera, I was going to give you a hint. This is the first example of organic development inside of HP Storage since the original EVA. The original EVA was a kick-ass product. I remember when we saw that, late 90s, early 2000s, we said, wow, this is ahead of its time. And then HP sort of squeezed the R&D investment and that's when they got into trouble. HP didn't make that mistake. You poured money into R&D, put resources in there. This is true organic development run by HP, not by the independent or smaller company that you acquired, so that's awesome. Congratulations on that. Yeah, absolutely, absolutely. Thank you, thank you. That's a very astute observation on what we did. So that was pretty long good observation. Yeah, so now we're here. You guys are out talking about the All Flash data center. Most companies aren't, well, Pure Storage will talk about the All Flash data center. They don't have anything else, but most large established companies aren't talking about the All Flash data center. They're sort of protecting the past a little bit from the future. They're sort of doing that balancing act. You guys don't have that strategy. You're going hard after All Flash. Talk about that. Yeah, so here's how we see from an opportunity standpoint and talking to customers standpoint is that less than two or 3% of the customers who have really made the transition from disk to flash. So we see that as a 98 or a 97% opportunity for just lying ahead of us. And this transition is real because Flash not only helps with the power saving and the acoustics and the data density, but the performance. So it's an, as customers go and look at the characteristics of Flash and look what the array can provide, they, we see a world in the next few years where the transformation is going to go from, maybe today I think I'll be optimistic by saying that most of the data center customers are probably at 2% of the Flash adoption that 98% of architecture can get there. And that's our addressable market right now, which is a fantastic opportunity for us from a revenue and growth standpoint. Well, and the question that we have in the crowd chat today is, okay, when, when is that going to happen? And my feeling is it's an economics at the point at which it crosses over, which is very close. So depending how you cut it raw, not yet, but if you throw in deduplication, you're there. I'll give an, I'll give a proof point is when, how the market adopts it, right? The last two quarters, the three power revenue has been, has been growing gang buster for us for several quarters now, right? Graining market shares and everything. In the last two quarters, we saw a very interesting trend in our install base. We sold more Flash than 15K RPM drives. So on a business which is growing in this fast, fast clip, you see that change happening on your high performance drives to the Flash and that will transfer. That's a leading indicator on what's happening and it's going to happen very fast. Well, you know, David Floyer and people go to the wiki, go to wiki bond.org and you'll find this piece on there, click the history in the wiki. In 2009, he predicted that by 2014, all Flash arrays would replace high spin speed, 15K RPM and people said at the time said 2000 for that, you're crazy. Right. He was, okay, pretty close. He was pretty close. Not bad. He's now saying that Flash, forgetting people costs, that Flash will be less expensive than spinning disk by next year. Now, what he's saying is the raw costs, you have to add in duplication. Absolutely. And you have to factor in the fact that you need less storage to service an application that's Flash, but if you do that, it's going to be cheaper. And then, his cherry on top is data sharing. Yep. Because you can now make developers more productive by giving them essentially a live copy of data that they can test and dev on. And that's going to, and you can't do that with spinning disk. You have to spin off copies and then you get copy creep and that kills productivity. So what are you seeing there? I think he's right on. He's right on on his predictions, both on the timing standpoint and also the reasons for why it's going to happen very fast. I couldn't agree with him more. In fact, in fact, his prediction is so right that next discover, I'm going to bring him to the blackjack table. We'll see. I mean, there's a very bold prediction now. And as well, in talking to some of your customers like we had Alcoa on the other day, they're not using Ddupe. They chose not to. They said, yeah, it's fine. We're making a lot of money off of this. We don't want to mess with it. We don't want to put that in. And you can understand that there's sensitivity there. So there's a maturity thing as well. So sometimes, you know, we analysts tend to get a little ahead of our skis, but the trend is very clear. Trend is very fast. And I think this is one of the trends which is going to accelerate very fast as well. So this is one of the fast moving trend. And I think people will make this transition at their own pace, but there is at least 98% of this customers they go in. And nobody wants to build a data center, another data, new data center. They want to optimize what they already have or even refine better. So I think it's the, they will use more features. The flash is also getting denser. I mean, we just, so we, six months ago, we were at 1.92 terabyte flash drive. Today we're at 3.84 terabyte. That happened like in six months. So, so even without the deduplications, compressions and everything, the density of the flash is also increasing. So you're tracking Moore's law, you're coming, the price is coming down faster than, than, than spinning this. Perfect storm. Just a matter of time. Perfect storm of economics. All right, Milan, we have to leave it there. Thanks very much for coming on theCUBE. It was great to hear in the story and congratulations on the success. Really appreciate it. Thank you very much. And we're glad to be here. All right, keep right there. But Jeff Frick and I will be back with our next guest. This is theCUBE or live from HP Discover 2015.