 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 2015. Everybody, my name is Dave Vellante. We just came off the keynotes. We heard Meg Whitman. We heard about HP's transformation. Martin Fink only talked very briefly about the machine. We heard about Gromit, a new open source project that HP is sponsoring, announcing, not HP Gromit, just Gromit. Greg Nunes is here, Vice President of Marketing, of HP Storage, Long Time, Cubulum, and Mike Abrams. And Mike is an IT practitioner at Alcoa. Gents, welcome to theCUBE. Yeah, thank you. Good to see you, Greg. Mike, first time on? Absolutely. We were just talking some football. Mike's from Pittsburgh, I'm from New England. Semi-rivals, I used to be a Pittsburgh fan when I was a kid because the Patriots were so bad, so I placed in my heart. So we know what it's like, Pittsburgh fans. You're intense, though. Yeah, absolutely. Are you as intense about IT as you are football, or almost? Yeah, I'm pretty passionate about season one. So you were telling me off camera, very interesting route to your current position. You came from the application group, and now you're in the infrastructure group. Yes. How'd that all come about? Really, just an opportunity for me at the corporation to do something different. Worked across our applications portfolios for roughly 16 years and kind of ran out applications to work on. So opportunity for growth for me to pick me up and make me really uncomfortable in a new role with a limited background and put me in the infrastructure space. And it's been a pretty interesting year and a half, but a lot of fun, a lot of learnings. Well, and I'm sure you as well bring a perspective of, hey guys, we shouldn't do that because the application guys are going to not be happy. What's there? Is that dynamic occur, and what's the? Yeah, without a doubt. What's it like? You know, one thing I learned at the applications group and the infrastructure groups are very different organizations. You tend to operate with your silo and your blinders on where all the applications guys are just thinking about building and deploying their apps. The infrastructure guys are really just focused on ensuring availability and delivery on the hardware, but very few make decisions that actually stop back and look from a perspective of, is the infrastructure decision I'm making in infrastructure going to impact the applications group? And likewise, when the application developers are building a platform, do they ever stop and look back as, hey, are the decisions I'm making going to impact the infrastructure? And you're very singly focused on your area and your role, and not necessarily thinking about what you do will impact the other side of the fence. So one of the learnings from my perspective, is establishing some new collaborations across the groups and trying to bring that application's perspective to infrastructure and that infrastructure perspective to applications. So Craig, this is relevant because you guys HP all in and all flash, and it's all about the applications, right? The storage used to be about how do you deploy it so the applications don't just slow down even more. That's just flipped, hasn't it? Absolutely, well I think the thing about flash and where we're at today, we've talked a lot about affordability to really bring it onto the radar, the conversation with folks. But I think, well in fact, Mike has just a wonderful story about the organizational productivity benefits that you pull out of a flash deployment, the problems you can solve without putting a bunch of your organization having to devote their lives to something you can solve in the technology. And the fact that that can make the application experience that much better, it seems like everyone wins. And I think the kind of where we're at today, kind of on this tipping point if you want to call it for this all flash data center is really helping folks understand that an all flash approach has an affordability aspect to it that we think we're looking at about a $1.50 per usable gigabyte with our compaction technologies. It's something you should be thinking about if you're thinking about 10 or 15 K disk, look at flash, but I think the real conversation and maybe the more important part of the conversation is what can it do for your organization? What can you do for your application users and kind of get that value out of it? So we love this conversation because as you know, we've been talking about it for years with our practitioner community. So Michael, but let's get through the cost knothole. So it used to be a bunch of flash into a disk array that was designed for spinning disk and eventually the controller bandwidth couldn't handle it. So that's cool and then all flash comes along but it's expensive, right? And then but the cost has been coming down. So what are you guys doing? Maybe describe briefly your application infrastructure and what are you doing with flash? I think, you know, at our journey, it's been an evolution and we started typically where I think the majority of organizations start and their consideration. You have a refresh opportunity come up with a highly critical application where you need to drive performance from a business perspective. And as you do that refresh, you look to put it on pretty much the fastest performing infrastructure you can acquire at the time. And you do that and you see the benefits and then you have a bunch of unintended benefits that you realize. You deploy your application and then you realize, wow, my backup window was just reduced by 90%. I wasn't thinking of buying flash to reduce my backup window. I was thinking to drive application performance and it gets you thinking. So you take a look as you have a lot of global application footprints and you have a lot of challenges of providing uptime and stability and performance while your Australia users are logging in to your app, post it out of your US data center and you take backups and performance, you know, degrade significantly, your help desk gets flooded with tickets and you say, hey, it's not a highly critical application but I'm going to take that because it's global and I'm going to put it on flash. I'm going to reduce my backup windows and provide a higher level of service to our global user base and in conjunction I'm going to reduce the workload on my help desk and my support organization by eliminating a lot of those performance tickets. It can get addictive as you start taking a look at every problem that you come up with and go, wow, I don't just throw that on the flash array. Flash heroine. It's an aspect of it, absolutely. So you have a significant portion of your infrastructure staff and today's day and age, it's very difficult to find a use case that you can't get the hardware in today's modern day that provides all the performance that you need. Your bottleneck now becomes your people. So you start looking at how do I free up, you know, resources in my organization to more value-added activities and you're fighting performance issues and you're fighting support issues and you take a look and say, you know, poor application code, badly custom developed code. Do I want my organization fighting that fire? Do I take that and put it on flash storage and free up, you know, 50 hours a week out of my storage group to focus on more value-added activity? So as you learn and as you grow and as you begin to use, you develop all these other use cases where you realize that the hardware can solve a lot of your modern day problems and have a lot of other unintended benefits that weren't necessarily what you originally went out to accomplish, but you're really glad that you have the learnings. And you used to be able to throw hardware, I mean, everybody threw hardware at the performance problems, but at some point it became cost prohibitive or you get diminishing returns. Are you saying you're seeing with flash that that's not the case? Without a doubt, you know, so you take a look at what HP did at Discover last year at $2, a usable gigabyte, and you started with now being able to acquire the flash storage at roughly what you were paying on fiber channel storage. So you take that investment that you knew you had to make and you invested at the same cost point, but you get significantly higher levels of service. In addition, if you have a little bit more of an expense in terms of going all flash, when you really start putting the business days together and look at, hey, if I can reduce 100 tickets a day out of my global help desk and the cost of every one of those tickets across three or four levels of support organizations, and you put a dollar amount on that and then you put a dollar amount and a value on time of your people, flash actually becomes far more cost affordable compared to some of the alternatives. So I wonder if we could break that down. So we're talking about, I don't know, is rough numbers, eight bucks a gigabyte raw, and then you reduce that down four or five X. So now that's how you get under two per gig. Is that roughly it? Yeah, call it about five raw. So it's five raw, five dollars a gig raw? Yeah, call it five raw. And so you're even less, you're even more conservative with your reduction. Yeah, four and change, then it's call it six protected with rate, and then a four to one compaction, you're down at a buck 50. Okay, and so you buy that, right? I mean, that's legitimate. You're seeing that kind of data reduction, I presume, right? That's not too radical. Is that right? Or does it kind of really workload dependent? Can you talk about that a little bit? It can be a little workload dependent. We're in our organization, we're not doing the de-dupe at this particular point. So we are running thin and we see a lot of benefits there. And now it's about our next evolution and going down the de-dupe route to achieve that lower number. And why aren't you doing de-dupe? That's just an internal choice, or? No, you look at in terms of its introduction and time to market, and sometimes you want to run out and be first, and sometimes you want to sit back and wait. So you're just watching, okay. You look at what you have on that array, and a lot of times, right, what you park there first were your highly business critical applications. And that's not necessarily your say, hey, I want to go start with de-dupe there. Yeah, let's not mess around with it because you're easily justifying the business value anyway, because you're talking about, we haven't really talked much about the productivity impacts. Well, let's talk about those. You mentioned- And I was going to say, Dave, I think the point being in terms of adoption of, I mean, on disk, we saw a relatively slow adoption of thin provisioning until folks kind of believed in the benefits. De-duplication, I think we see a bit of that. And for folks using it, we're seeing between thin and de-dupe above a four to one compaction rate. But I think the cool thing that happened was when we talked about two bucks per usable gig, we had the conversation. And while Mike is not using that, the organizational productivity benefits he's got, he's over the moon with that. And then when the team is ready to go de-dupe or technology is beyond that, that's just gravy, right? And so that's kind of the point, is get folks to consider flash versus disk and think through not just the TCO, but the organizational benefits. Yeah, and it's really been amazing. I mean, I remember last year, right around this time, I was speaking to a lot of different customers on roadshows and things, and very few had installed an all flash array. I mean, it was less than 10% that are really all in, and that's changed dramatically now. And we've been talking about the productivity benefits for years. David Floyer wrote a piece three years ago about the impact on productivity. So I wanted to ask you specifically, Mike, about what we're referring to now as data sharing. When you have a copy of data on spinning disk, you're making copies, you're populating data warehouses, you're doing test and dev, and so you're spinning up copies and you get copy creep. And the currency of that data is not live, right? With flash, we've talked to a number of customers that are able to serve up copies from that same physical infrastructure using live data, essentially, current data. Are you doing that with your test and dev? Are you seeing that? And what kind of impacts is that having? We see a couple of different use cases of benefits there when you talk from a pure infrastructure perspective of building, test, and dev. You obviously cut a lot of time in terms of your ability to copy that data. You go from latency on the storage to challenging your network at that point. But ultimately, you can significantly reduce the window that it takes for you to spin up those non-production environments and support. You see a lot of benefit from that perspective and that really helps to pre-up your infrastructure organization as you reduce the level of effort it takes to look in a large shop where you have a requirement of having a QA and a dev for every application that you run and then you look at that holistically, you really can add or save a significant people time of what it takes to construct those environments. And to be clear in Mike's environment, where you have active copies on your production volume, I mean, to your point, Flash has a huge economic benefit. I think your environment, if I have it right, you're actually taking those active copies off of a replica, off of a replicated environment. Am I right there? Yeah, and in certain instances, so we have a various DR that we provide from our applications in a one and three and seven and 30 day and the one we do active RCIP synchronization to a hotsite in our alternate data center and we'll build the copies off of that. And in addition, a lot of your lower level of service that you want to provide, then it becomes your ability to be able to push those to your backups, which you significantly improve. Yeah, yeah, so I think the key is for anyone who is, doesn't have a setup quite like Mike's and they are driving those copies off that primary production volume. Rarely would you run your active copies via snapshot off of that. You're going to take a physical copy away. There's huge capacity savings with Flash because you have the benefit of going right from. Yeah, so we just try to look at this from a top down perspective. You require less capacity, presumably, than you would with spinning disk. You can data reduce it if you choose to do so. You've got performance benefits and you've got other data sharing benefits that drive the cost down even further. And presumably as maturity improves, you're going to turn those knobs even harder over time. So my question is, do you buy into the All Flash data center? Mike and I had a conversation with this before we came on. It's a radical concept, right? I mean, I got the All Flash laptop. Actually, I don't have it. I want you to answer truthfully. Yeah, so let's have that conversation. What concerns you about the All Flash data centers? Just things don't happen that way? Well, so there's first and foremost, you see the reduction in cost in All Flash and that's fantastic. And you also respect or expect a corresponding reduction in your spinning media. If both the costs, you see flash dropping and the cost of your disk holding true, then I would absolutely see that happening by over the course of time. We all know that that media is going to become cheaper as well. And you look at specific use cases in your data center and you have a lot of, one, your data always grows. I never see it get, it never shrinks. Archive is sometimes challenging and depending on what type of products and markets you operate in, data retention requirements go out 50, 100 years. And archive becomes a challenge so you keep the data inside the application. From our user base, 99% of our queries had active year plus one. And when you have years three through 15 in the application it's used 1% of the time, once a quarter or very rarely. So you have with modern day architectures, the ability to segregate all that data with partitioning. And if I can get All Flash at $1.50 a gig or spending media at 75 cents a gig when I know that I don't have to provide that higher level of service for all our data. I have less of an investment in the data that I'm not using in that use case and then I capture that money I'd save and buy Flash for other production applications. How about if we took the archive piece out? Because look, I mean, we've still got tape, let's get serious, we still have tape for tape here. Flash and tape, it's coming. We have, you know, 7K RPM drives for exactly what you're talking about or tape. But for active data, for the majority of the IO you're doing, I think our premise to be clear when we talk all Flash data center, we're talking about the active processing, going to Flash, being ready for that now, tally up your IOPS. We heard from Salesforce.com marketing cloud yesterday, 80% of their IOPS today, today are Flash. Yeah, so you're saying all active data will be on Flash. That's what you mean by the all Flash data center. David Fleurer, David Fleurer on the other hand says that basically Flash will be cheaper than all spinning disk by next year, which is a very radical prediction. Now he's assuming- The copy benefit. That data reduction adoption is there. So maybe take some time. He's also assuming some kind of data sharing, enlightening practitioners say, oh yeah, okay. But that takes time to absorb that type of thing. But like you're saying, if I understand what you're saying, you agree that all the active data is going to be on Flash, but then there's another tier. So you're talking about the traditional storage hierarchy changing and do you use tape? No, no tape, not even for deep offline last resort. Only time we'll engage in a tape-based backup is when there's a legal requirement to hold data beyond our backup policy. Yeah, so not for backup. I realize you're not using tape for backup, but maybe for long-term retention, you might. Yeah. And not for archiving. Correct. I have another consideration for you. And this is kind of, you know me, right? Sort of, it's a little more emotional. But here's the thing. As consumers, we are all, whether you know it or not, completely addicted to Flash. Totally. It's in our phones, it's in our laptops, it's in our notebooks, it's everywhere, right? And if you took Flash away, our digital world would slow to a crawl. And it's kind of a terrifying thought as a consumer, like I'm not sure I could function in that world. I couldn't. And you kind of hinted at it, Mike, when you said it's kind of addictive as you move things to Flash and the benefits you get, you kind of want more of that. And the notion that as you move active data to Flash, you're going to see benefits you didn't anticipate. Your users are going to love that and you're going to find ways to take advantage. And I am here telling you that the all Flash data center, I mean, when you start down that road, you're going to be there before you know it. It's like when you can't get on the network and you're spinning, you're like, huh, can't get on the network. And I think Flash is taking a similar path. But Mike, pragmatic approach, I'll give you the last word. Where's Alcoa sort of going in this direction it's journey, it's digital enterprise, it's idea, economy, what's next for you guys? So the one thing that you learn is as soon as you establish a new performance level, what's your new benchmark? And that always becomes just the right of passage or entry at that particular point. Table stakes. Once that's established, you're there. Everything gets measured up to that. So if you look at what benchmarks you established, the productivity you create inside your organization, absolutely more and more and more workloads, you move to all Flash. And that absolutely is the direction that we've been taking. All of our last storage additions in terms of our data centers, we've been putting on Flash, we've been putting non-traditional applications in Flash and garnering the benefits from that. We've put the problematic applications on Flash and free up the people in the shop and we've seen the benefits there as well. You know, so I certainly wouldn't argue in all Flash data center is not in the too distant future and particularly if you talk about active data, I absolutely could see that as the cost continues to decrease in your constrained shift and they move away from hardware and become more people oriented, you always look where you could solve the problems and it's a toll that you have in your pocket to solve a lot of problems and once you start using the toll, you typically never put it away. All right, gentlemen, thank you. Good discussion. All Flash data center, all Flash ish, data center, Flash, good for what ails you and getting less and less expensive. All right, keep it right there, everybody. We'll be back with our next guest. This is theCUBE, we're live from HP Discover 2015.