>> Announcer: Live from San Francisco, it's theCUBE. Covering Pure Accelerate 2017. Brought to you by Pure Storage. >> Welcome back to Pure Accelerate. We're here at Pier 70 in San Francisco, and this is theCUBE, the leader in live tech coverage. I'm Dave Vellante with my co-host David Floyer. Matt Kixmoeller is here, he's the Vice President of Product and Solutions at Pure Storage. Kix, welcome to theCUBE. >> Thanks for having me. My first time on theCUBe, I'm honored. >> That's awesome, well, we're honored to have you. Got to have a nickname on theCUBE. We had Dietz on earlier, Stu had to leave. You can call me V, if you want. You really don't have a nickname; we call him Floyer. (laughter) >> All right. >> So anyway, great job today on stage. You got a really engaged audience. You guys have a lot of fun. The orange shoes are cool. How do you feel? >> I feel great. You know, as we said today, this is the biggest year we've ever had in innovation at Pure, and it was fun to really take the focus back to software this release. You know, we spent the last year bringing out our next-gen cloud era all flash platforms, between FlashBlade and FlashArrayX, and this was an opportunity to really flex our muscles around software, flex our muscles around IoT and AI and that as well. So, it was a fun set of releases. >> Well, it's been interesting to sort of watch you guys and watch your product strategy evolve. And of course, coincident to that is your TAM expands. All right, so it started in the sort of you know, lower end of the spectrum, and then it went into the 20s and now it's in the 30s, and I was saying to David it used to be, well I buy EMC for block and I buy NetApp for file, and you guys are challenging that convention. >> Matt: Yeah. >> Maybe talk a little bit about your strategy and how your penetrating now new markets. >> Yeah, we think about our market opportunity in three buckets. So first off, we go after the top 500 cloud providers, and we see one of our biggest segments is really cloud providers and we see them increasingly not really looking at legacy options for storage. You know, they want a modern storage fabric, and part of why we're so excited in particular about the work we've done around NVMe is we feel like it helps us go after some of the more server DAS-centric workloads of the past, or of the next gen workloads, and we can talk a little bit more about that. A second key area that we're focusing on is really going after next generation data-driven applications, and AI, ML, all these areas are really driving amazing storage growth. It's even, I think, surprised us how quickly it's come up, and you had folks on theCUBE earlier today talking about FlashBlade, but one of the threads that units a lot of the next gen applications is they're designed to be scale-out and they're designed to really need a lot of parallelism from storage. And so what we're doing with FlashBlade is really designing a storage platform that's kind of parallel from the start, and can deliver that massive concurrency that you just can't get from a lot of legacy providers. And then yeah, I think the third thing we're obviously excited about is going in and ripping out the spinning rust of the past. We've made a lot of innuendos at this conference and how we're in this classic rusting building and maybe it's a nice metaphor for some of that-- >> Tear it down! (laughs) >> But yeah. But we are we're helping liberate the rest of the world, and I think one of the things that we're excited about today was to announce Purity Active Cluster. That's been that top of the reliability hill feature when people want metroclustered applications, active-active in two data centers, that's about as reliable as it gets, and that was a feature that we didn't have in FlashArray until now, and so we're excited to have that final area to go in and help liberate. >> Yeah, so it's not just the disk spinning rust replacement, it's, you talked this morning about SRDF, I remember well in the early 1990s when SRDF came out, it was game changing and it obviously has driven a lot of revenue for EMC, now Dell EMC, it helped a lot of customers, but there's no question it was the mother of all complexity and cost. So talk a little bit more about how you guys are going to approach that problem. >> Yeah, I mean, I think if you look at a lot of what we announced today, there continues to be a thread of simplicity throughout everything. You know, it's fun, I was employee number six up here, I've been in on the adventure from day one, right, and we always had a fundamental belief in simplicity. But as we started to shift products and started to get customer feedback, there was like this lightning rod within our team all throughout engineering where people really understood the power of simplicity. And it went from a belief to a religion, I would almost say. And we've just always tried to do that with new feature we come out with, and this felt like an area where there was such a vacuum of simplicity that there was a huge opportunity to rethink things. And so, with this feature it's totally built in, it's totally integrated, you could easily just stretch a volume across now two sites. And one of the problems we went to go solve was the third site mediator problem where you always need a third site witness in a stretch cluster to determine if there is a failure, who's the surviving side that you want to have actually process the application IO. And so we're delivering that as a service, as a SaaS service from our Pure One infrastructure, so it's just one more way that we take one more step and one more pain of the infrastructure away. >> So I'd like to drill in a little bit on the NVMe side of this. >> Matt: All right. >> We've done some research on the architecture which we think is coming up, which we're calling unigrip because it allows this very even access to data at very low latencies across there. And really, we'll start in our view, a different sort of applications, really very very different where you can combine legacy state applications with the AI applications and other things like that. How are you going to bring that to market? Who are you selling that to? >> Yeah, I mean, we're super excited about this transition, NVMe and we're trying to take a real leadership role here. And so much of it reminds us actually of the early days of Pure. When we started Pure, flash was expensive, it was exotic, you had a bunch of people trying to make it this 1% technology, and our whole idea was look, let's not make it a Ferrari, let's democratize it for all and we think everybody deserves flash. And we did a bunch of work to try to mainstream it. And we're trying to take a very similar approach with NVMe, where a lot of the early folks who approached NVMe built very specialized appliances, did exotic things. And our view is it should be mainstream. All flash arrays should be built on NVMe. And the real advantage is something you hinted at. It's just massively parallel. And so here you have flash, this inherently parallel medium on its own and we're talking through it through these legacy SCSI protocols that have been around forever. NVMe is a huge opportunity to open that up. But we had an initial insight, I believe, where when we approached this we didn't just say look, we should get an NVMe SSD. We realized that that whole architecture has to be optimized from software to hardware, and so we forgoed or forwent the SSD form factor. We built our own direct flash module, and the real magic of how we've approached this is not only shipping a device that's massively parallel, but building a bunch of software within Purity that knows how to take advantage of that, and brings all the flash management up to the software tier, so we can kind of take advantage of it end to end. And so, these are things we just don't see our competitors in the market doing right now. Maybe one more comment on your parallelism. I mean, I think you're right in that if you look at a wide range of next generation web-scale applications, whether they be more classic NoSQL databases on through analytics, on through to AI and ML. AI and ML are kind of maybe the most extreme examples, but they're all far more parallel scale out applications than we were used to before. And so they thrive in environments where you have storage that can marry that model. And what we're finding in particular in the AI world is that we're not up against other storage vendors. I mean, the alternative really is to go get a bunch of white-box DAS and build your own storage layer and maybe use some open source stuff, but that's cumbersome and that has all the issues that everyone's aware of, right? So we believe that as a commercialized product we have something pretty unique to offer these markets and it's been exciting to see it even push us. One of the things I think we surprised people with today was making FlashBlade 5x bigger. You know, we announced it last year, people thought it was pretty big and pretty fast to begin with, but it was these use cases and the early adopters that pushed us to make it larger. We saw people in the early adopter phase of FlashBlade buy in and deploy at much bigger scale than we were expecting. We were kind of used to our experience with FlashArray where people sort of started small, they got to use the technology, then they kind of grew. But I guess you don't do big data on a small scale. (Laughter) So people dived in. >> So I want to ask you about this whole big data, because it's probably the first time we've even used that term today. It's amazing how fast that came and went, even though big data's now mainstream. But, and you said, you made the point, Matt, that not a lot of storage competitors are going after that. Well, you'd think big data, storage, they would fit. But I think a lot of the competitors realized well, there's not a lot of money to be made there. And now it's just hitting its best stride. Here's my question. If you look at Hortonworks and Cloudera in particular, you're starting to see the cloud guys, Amazon with its data pipeline, certainly Google and Microsoft, are picking up a lot of action in the cloud with a full as-is service of the data pipeline. What do you see, and it's affecting some of the on-prem activity, what are you seeing with regard to cloud versus on-prem, and how does that affect your business? >> Yeah, I think you're right in the sense that if you looked at how you could have deployed big data technologies before, I think that there are basically two ways to do it. People that did it in the cloud, or they did it on-prem with white-box DAS, and they've got servers and put disks inside. So much of the first generation of big data was basically driven on Hadoop, which fairly low-cost and fairly focused at streaming workloads where you had this, frankly not much performance profile or need for performance on disk, and so what we found in the early days was, hey if you tried to put flash underneath it, didn't help that much. >> Dave: Didn't do much for it, right. But the thing that's changing now is people want to move away from those slow batch queries to much more interactive analysis, much more real time, and so Hadoop's given way to Spark, and so that's changed that discussion quite a bit. Back to the discussion though, around on-prem versus the cloud, I think this is an area where as people get more and more invested in their data, they're understanding it's a key control point. And so if I get all my data into one cloud provider, it's pretty hard to get it out of there. This is core to my business. Do I want that level of lock-in? Also, can I do better with my own dedicated solutions? And what we've found is that when we can bring FlashBlade to bear these big data workloads, we can outperform what people can do in the cloud handily, at a lower cost. And so there's a proclivity to want to own your own destiny, own your own infrastructure, and the ability for us to deliver a higher performance for a lower cost in the cloud we think is a pretty good connection. >> And of course, complexity is hurt, it isn't hurt, I mean, the market's growing very nicely, but it's actually hurt a lot of the practitioners' ability to absorb technology. I suppose Pure and its insane focus on simplicity helps a little bit, as does Spark, sort of simplify the whole Hadoop thing, but you've still got, you need a lot of smart people to make this stuff work. So it's going to be interesting to see, but what I'm hearing from you is you don't have a lot of storage competitors going hard after this. And so the guys that have done really well with Hadoop that have on-prem infrastructure you would think would be picking this up quite rapidly. Well, and look, we're having discussions with all of the Hadoop providers as well, because if we can help them deliver a higher customer satisfaction and a better outcome, it's upside for them as well. They don't want to be storage companies. >> Well, they need help, I mean the irony is that Cloudera is in the cloud era, and the cloud is eating away at its base, so they need somebody who's going to help them simplify, I mean, they're a software company, help us simplify the on-prem infrastructure. >> One of the things you said earlier that I think has been an additional learning for us, and FlashBlade as well, when we went into the FlashBlade experience, we kind of expected that people would buy and all they would care about is performance. And so we asked ourselves, well how much does this user base really care about simplicity? We found the total opposite to be true. Most of who we're selling FlashBlade to are not IT folk. They're data scientists, they're engineers, they're creatives, they're a line of business people. And they want nothing to do with managing infrastructure. And so the simplicity, oftentimes we're replacing what would have been racks and racks of disk that they didn't want to deal with to begin with. And so the simplicity value prop, shockingly, is actually more important, we're finding, for FlashBlade even than FlashArray. >> Makes a lot of, we have a saying in theCUBE that data is the new development kit. 'Cause it's like you say, it's data engineers, it's data science, even application developers are starting with the data, and so, and complexity has choked that whole industry, and so that's excellent. Okay, are you? >> Oh yeah, I was going to ask. One of the things you were saying very clearly here is that the drive of getting data up to the cloud to do this AI, or up to anywhere to do the processing, to create the models, is going to have to be ameliorated by reduction of that data. By reduction, I mean turning that data into informational tags or whatever it is as it's going up the line, very close to where the data is. >> Dave: I call it the needles in the hay stack. >> Yeah, extract the needles very early on. So can you talk a little bit more about what your vision is there, how are you going to do that, who are you partnering with to do that? >> Yeah, so I think that you hit on a very important problem, and I think everybody is starting to finally internalize how much faster devices and machines can generate data than humans. (Laughter) And so we're used to this human era of cognition of data creation, but this asymptote is happening. And, you know, I think it's becoming quite obvious that basically machines have the potential to generate data much faster than it can be stored, used, and especially sent back to the cloud. And so you need some level of local processing to analyze it, to send back more, you know, kind of per that metadata. The other challenge is that many of the use cases that people want to use at the edge are latency sensitive, and so you can't take the time to think about it, send it all back, think about it, send it back again-- >> Dave: Ogle it. >> And do some realtime control thing, right? My favorite anecdote that proves this is some of Amazon's infrastructure, where they build out dedicated data centers within their distribution facilities because they need to be able to realtime analyze the video feeds of everything that's going on, make decisions, right? And so if they can't send all the data to their cloud, they have to build they're own data center-- >> Dave: Nobody can! >> Inside there. (laughter) And so it's just indicative of a broader solution there, right? You'll see a demo that we're going to be doing tomorrow where we're doing a great coprocessing app where we're kind of collecting a bunch of data here at the show, analyzing it, and then sending part of it up to the cloud and partnering with Google to analyze it there, and showcasing an example use case of this. And so we think it's an area that's going to be important. Part of that also brings us to what we've done with our Purity run. So one of the things we've announced today was opening up our Purity platform to third party code, to developers. And we see a number of use cases for this. Many of our cloud customers have asked for this, where they want to kind of tie the storage more directly into their application, but the other use case we see is the edge. Where, if we can deploy a local Pure device on your oil rig, in your plane, in your factory, whatever, and have that processing capability happen there, and then to have that summarize the data and be able to send it back, that provides more of an all-in-one solution for that. And so, you know, we don't have dedicated products in this space yet, but this is our way of opening up the platform to be able to see how people develop on that how how they can start taking advantage of that. >> Okay so, we got to wrap, but you were telling us you were employee number six-- >> Matt: Yep! >> So that's quite a ride. I mean, so many companies just don't get to reach escape velocity, to use that term. You guys did. What's next for you, where do you want to take this thing? >> Yeah, I think we're all extraordinarily excited here at Pure. I mean, so much of this first generation of Pure's growth has been reshaping the existing storage environment. And, you know, we feel like we're through that mission. Yes, okay, only 20% or so of enterprise storage is flash, but the writing's on the wall, we're delivering the products. That is momentum now, right? >> Dave: Right. >> And so so much of our next generation of innovation is going after these new data-driven use cases, helping cloud providers, just going after what's next. And that opens up a much broader definition of what you can be as a data company. You know, we kind of stopped referring to ourselves as a storage company, we're going to have to get storage out of the name at some point, but you know, going after the broader problems around data is a much more exciting mission that we think powers the next decade, so, lots to do. >> Great, right, Kix, thanks very much for coming to theCUBE. >> Matt: Thank you guys! >> It's great to have you. >> Floyer: Thank you. >> Matt: Appreciate it. All right, keep right there, buddy, we'll be back to wrap up right after this short break. This is theCUBE, we're live from Pure Accelerate 2017. Right back. (upbeat electronic melody)