 From Austin, Texas, it's theCUBE, covering Pure Storage Accelerate 2019. Brought to you by Pure Storage. Welcome to theCUBE, the leader in live tech coverage, covering Pure Accelerate 2019. Lisa Martin with Dave Vellante in Austin, Texas this year. Pleased to welcome a couple of guests to the program. Please meet Charlie Boyle, VP and GM of DGX Systems at NVIDIA. Hey Charlie, welcome back to theCUBE. We've been on a long time ago, and we have Brian Schwartz, VP of Product Management and Development at Pure. Brian, welcome. Thanks for having me. So here we are, day one of the event. Lots of news this morning. Pure is just about to celebrate its 10th anniversary, a lot of innovation in 10 years. NVIDIA partnerships about two-ish, two and a half years old or so. Brian, let's start with you. Give us a little bit of an overview about where Pure and NVIDIA are, and then let's dig into this news about the AI data hub. Cool. It's been a good partnership for a couple years now, and it really was born out of work with mutual customers. We brought out the FlashBlade product. Obviously NVIDIA was in the market with DGX's for AI, and we really started to see overlap in a bunch of initial deployments. And we really realized that there was a lot of wisdom to be gained of some of these early AI deployments of capturing some of that knowledge and wisdom from those early practitioners and being able to share it with the wider community. So that's really kind of where the partnership was born. Going for a couple years now, got a couple chapters behind us and many more in the future. And obviously the AI data hub is the piece that we really talked about at this year's Accelerate. Yeah, Aries about been in the market for what, about a year and a half or so? Almost two years now. Almost two years. All right, tell us a little bit about the adoption, what customers are able to do with this AI-ready infrastructure? Yeah, and as pointed out, the reason we started, the partnership was our early customers that were buying DGX product from us and they were buying Pure Storage, both leaders in high performance. And as they were trying to put them together, they were like, how should we do this? What's the optimal settings? They'd been using storage for years. AI was kind of new to them and they needed that recipe. So that's the early customer experiences turned into Aries' solution. And the whole point of it is to simplify AI. AI sounds kind of scary to a lot of folks and the data scientists really just need to be productive. They don't care about infrastructure but IT has to support this. So IT was very familiar with Pure Storage. They'd used them for years for high performance data and as they brought in the NVIDIA compute to work with that, having a solution that we both supported was super important to the IT practitioners because they knew it worked, they knew we both supported it, we stood behind it and they could get up and running in a matter of days or a week versus six to nine months if they built it themselves. When you look at companies that you talk to, customers, let's narrow it down to those that have data scientists, at least one data scientist and you ask them where they are in their maturity model. Kind of, if one is planning, two is early, three is they got multiple use cases and four is they're enterprise-wide. How do you see the landscape? Are you seeing pretty aggressive adoption in those as I couched it or is it still early? So every customer's at a different point. So there's definitely a lot of people that are still early but we've seen a lot of production use cases. Everyone talks about self-driving cars but there's a lot behind that but real-world use cases say medical's got a ton. We've got partner companies that are looking at reconstruction of MRIs and CT scans cutting the scan time down by 75%. That's real patient outcome. We've got industrial inspection, we're in Texas. People fly drones around and have AI models that are built in their data center on the drones and the field operators get to reprogram the drones based on what they see and what is happening real-time and then it retrains every night. So depending on the industry really depends on where people are in their maturity curve but really our message out to the enterprises is to start now. Whether you've got one data scientist, you've got some community data scientist, there's no reason to wait on AI because there's a use case that will work somewhere in your enterprise. So what are the key considerations to getting started? What would you say? So one thing I would say is to your stages of maturity any good investment is done through some creation of business value and an understanding of what problem you're trying to solve and making sure it's a compelling problem is an important one and some industries are farther along. One of the ones that most everybody's familiar with is the tech industry itself. Every recommendation engine you've probably ever seen on the internet is backed by some form of AI behind it because they want it to be super fast and customized to you as a user. So I think understanding the business value creation problem is a really important step of it and many people go through an early stage of experimentation, data modeling, really kind of, I'll say a prototyping stage before they go into a mass production use case. It's a very classic IT adoption curve. Just to add a comment to the earlier kind of trend is it's a mega trend. Yes, not everybody is doing it in massive wide scale production today. There are some industries that are farther ahead. If you look forward over the next 15 to 20 years there's a massive amount of AI coming and it is a new form of computing the GPU driven computing. And the whole point about AI is getting the ingredients right to have this new set of infrastructure, have storage network compute and the software stack all kind of packaged together to make it easier to adopt to allow people to adopt it faster because some industries are far along and others are still in the earlier stages. Right, so how do you help for those customers and industries that aren't self-driving cars or the drones that you talked about where the use case we all understand it and are excited about it but for other customers in different industries how do you help them even understand the AI pipeline and where do they start? I'm sure that varies by customer but give me some favorite examples. It does vary a lot but the key point is start an AI project that you have a desired outcome from not everything's going to be successful but AI projects aren't something that it's not a six month IT project or a big CRM refresh it's something that you could take one of our classes that we have we do a lot of end user customer training with our deep learning institute you can take a half day class and actually do a deep learning project that day and so a lot of it is understanding your data and that's where Pure and the data hub comes in understanding the data that you have and then formulating a question like what could I do if I knew this thing? That's all about AI and deep learning it's coming up with insights that aren't natural when you just stare at the data how can the system understand what you want and then what are the things that you didn't expect to find that AI is showing you about your data and that's really a lot of where the business value comes and how do you know more about your customer how do you help that customer better? AI can unlock things that you may not have pondered yourself. Yeah, the other thing I'm a huge fan of analogies when you're trying to describe a new concept to people and there's a good analogy about AI data pipelines that predates AI around data warehousing like there's been an industry around extract, transform and load ETL systems for a very long period of time it's a very common thing for many people in the IT industry and I do think there's when you think about a pipeline an AI pipeline there's an analogy there which you have data coming in ingress data you're cleansing it, you're cleaning it you're essentially trying to get some value out of it how you do that in AI is quite a bit different because it's GPUs and you are looking for turning unstructured data into more structured data it's a little different than data warehousing traditionally was running reports but there's a big analogy I think to be used about a pipeline that is familiar to people as a way to understand the new concept. So that's good I like the pipeline concept one of the counters to that would be that when you think about ETL it's complicated process enterprise data warehouses that were cumbersome do you feel like automation in the AI pipeline when we look back 10 years from now we'll have maybe better things to say than we do about EDW and ETL. And I think one of the things that we've seen obviously we've done a ton of work in traditional AI but we've also done a lot in accelerated machine learning because that's a little closer to your traditional data analytics and one of the biggest kind of aha moments that I've seen customers in the past year so it's just how quickly by using GPU computing they can actually look at their data do something useful with it and then move on to the next thing so that rapid experimentation is all what AI is about it's not a one and done thing and lots of people think oh I have to have a recommender engine and then I'm done no you have to keep retraining it you know day in and day out so that it gets better and that's before you had accelerated AI pipelines before you had accelerated data pipelines that we've been doing with GPUs it just took too long so people didn't run those experiments now we're seeing people exploring more trying different things because when your experiment takes 10 minutes two minutes versus two days or 10 days you can try it you know your cycle time shorter businesses can do more and sure you're going to discard a lot of results but you're going to find those hidden gems that weren't possible before because you just didn't have the time to do it isn't a key operationalizing it as well I mean again one of the challenges with the analogy that you gave on EDW was fine reporting you could operationalize it for reporting but the use cases weren't as rich and robust and I feel as though machine intelligence is I mean you're not going to help but run into it it's going to be part of your everyday life your thoughts on it it's definitely part of our everyday lives you know when you talk about you know consumer applications of everything we all use AI everyday you just don't know it it's you know the voice recognition system getting your answer right the first time you know there's huge investments in natural language speech right now you know to the point that you can ask your phone a question it's going through searching the web for you getting the right answer combining that answer reading it back to you and giving you the web page all in less than a second you know before you know that'd be like you talk to an IVR system wait then you go to an operator you know now people are getting you know such a better user experience out of AI back systems that you know over the next few years I think you know end users will start preferring to deal with those base systems you know rather than waiting online for a human because it'll just get it right it'll get you the answer you need and you're done you save time the company save time and you've got a better outcome so there's some definitely some barriers to adoption skills is one obvious one the other and I wonder if Pure and Nvidia have you know attacked this problem I'm sure you have but I'd like some color on it is traditional companies which a lot of your customers their data is in pockets it's not at the core you look at the AI leaders you know the big five data they're data companies it's at the core they're applying machine intelligence to that data how has this modern storage that we heard about this morning affected that customer's abilities to really put data at their core you know it's a great question Dave and I think one of the real opportunities particularly with flash is to consolidate data into a smaller number of larger you know kind of islands of data because that's where you can really drive the insights historically in a disk driven world you would never try to consolidate your data because there was too many bad performance implications of trying to do that so people had all these pockets and even if you could you probably wouldn't actually want to put the data on the same system at the same time the difference with flash is there's so much performance at the core of it at the foundation of it so the concept of having a very large scale system like the 150 blade system we announced this morning it is a way to put a lot of data and be able to access it and to Charlie's point a lot of people they're doing constant experimentation and modeling of the data you don't know how the data is going to be consumed and you need a very fast kind of wide platform to do that which is why it's been a good fit for us to work together now follow up on that data by its very nature however Brian is distributed and we heard this morning is you're attacking that problem through you know an API framework that you don't care where it is cloud, on-prem, hybrid, edge at some point in time your thoughts on that well and again the data to be used for AI I wouldn't say it's going to be every single piece of data inside an organization is going to be put into the AI pipeline in a lot of cases you can break it down again to what is the problem I'm trying to solve the business value and what is the type of data that's going to be the best fit for it there are a lot of common patterns for consumption in AI speech recognition image recognition places where you have a lot of unstructured data or it's unstructured to a computer it's not unstructured to you when you look at a picture you see a lot of things in it that a computer can't see right, because you recognize what the patterns are and the whole point about AI is it's going to help us get structure out of these unstructured data sets so the computer can recognize more things you know the speech and emotions that we as humans just take for granted it's about having computers being able to process and respond to that in a way that they're not really capable of doing today hot dog, not a hot dog you know, the Silicon Valley fans it's the streetlight which one of these is not a streetlight so you prove you're not a bot I want to ask you about distributed environments you know, customers have so much choice for everything these days on-prem, hosted, SAS, public cloud what are some of the trends that you're seeing I always thought that to really be able to extract a tremendous amount of value from data and to deliver AI from it you needed the cloud because you needed a massive volumes of data the purest legacy of on-prem what are some of the things that you're seeing there and how is NVIDIA and Pure coming together to help customers wherever this data is to really drive value, business value from these workloads I'll have two quick comments then I'll turn it over to Charlie so one is we get asked this question a lot like where should I run my AI the first thing I always tell people is where's your data gravity moving, these data sets are very large tens of terabytes, hundreds of terabytes petabytes of data moving very large sets of data is actually still a hard challenge today so running your AI where your data is being generated is a good first principle and for a lot of folks they still have a lot on-premise data that's where their systems are they're generating these systems or it's a consolidation point from the edge or other opportunities to run it there so that's where your data is run your AI there the second thing is about giving people flexibility we've both made pretty big investments in the world of containerized software applications those things are things that can run on-prem or in the cloud so trying to use a consistent set of infrastructure and software and tooling that allows people to migrate and change over time I think is an important strategy not only for us but also for the end users it gives them flexibility so ideally on-prem versus cloud implementations shouldn't be different it'd be great if they'd be identical but are they today? so at the lowest level there's always technical differences but at the layers that customers are using it we run one software stack no matter where you're running so if it's on one of our combined ARI systems whether it's in a cloud provider it's the same NVIDIA software stack from our lowest end consumer grade GPUs to the big 350 pound DGX2 you see back there we've got one software stack that runs everywhere and one that Ryan was making it's really run AI where your data is and while a lot of people if you are a cloud native company if you started that way I'm going to tell you to run on the cloud all day long but most enterprises they're you know some of their most valuable data is still sitting on premise they've got decades of customer experience they've got decades of product information that's all running in systems on-prem you know when you look at speech speech is the biggest thing you know they've got years of call center data that's all sitting in some offline record what am I going to do with that that stuff's not in the cloud and so you want to move the processing to that because it's impossible to move that data somewhere else and transform it because you're only going to actually use a small fraction of that data to produce your model but at the same time you don't want to spend a year moving that data somewhere to process it you know back the truck up put some DGXs in front of it and you're good to go someone's going to beat you to finding those insights right so there is no time so you have another question I have the last question so you go ahead so in video you got to be Switzerland in this game so I'm not going to ask you this question but Brian I will ask you why is Pure different I know you were first you raced out you got the press release out first but now that you've been in the market for a while what are Pure's competitive differentiators you know there's really two I would net it out for FlashBlade on why we think it's a great fit for an AI use case one is the flexibility of the performance we call it multi-dimensional performance small files, large files metadata intensive workloads FlashBlade can do them all it's a ground up design it's super flexible on performance and but also more importantly I would argue simplicity is a real hallmark of who we are it's part of the modern date experience that we were talking about this morning you can think about these systems they are miniaturized supercomputers and yes you could always build a supercomputer people have been doing it for decades you use PhDs to do it and like most people don't want to have people focused on that level of infrastructure so we've tried to give incredible kind of capabilities in a really simple to consume platform I joke with people we have storage PhDs like literally people at PhDs for storage so customers don't have to and Charlie, feel free to chime in on your favorite child if you want you know a lot of it comes from our customers that's how we first started with Pure it was our joint customer saying we need this stuff to work really fast they're making a massive investment with us in compute and so if you're going to run those systems at 100% you need storage that can feed them Pure was our first in there they're our longest partner in this space and it's really our joint customers that put us together and to some extent yes we are Switzerland we love all of our partners but we do incredible work with these guys all up and down the stack and that's the point to make it simple if the customer has data we want it to be as simple as possible for them to run AI whether it's with my stuff, with our cloud stuff with all of our partners but having that deep level of integration and having some of the same shared beliefs to just make stuff simple so people can actually get value out of the data have IT get out of the way so data scientists can just get their work done you know that's what's really powerful about the partnership and I imagine, I know we're out of time but I imagine to be able to do this at the accelerated pace, accelerated I'm going to say pun intended it wasn't but cultural fit has to be pretty aligned we know Pure's culture is bold last question Brian that's what we bring it home here talk to us about how the cultural cultures of Pure and NVIDIA are stars aligning to be able to enable how quickly you guys are developing together you know we mentioned the simplicity piece a bit the other piece that I think has been a really strong cultural fit between the companies is just the sheer desire to innovate and change the world to be a better place you know our hallmark, our mission is make the world a better place with data and it really fits with the level of innovation that obviously NVIDIA does so two Silicon Valley companies with wicked smart folks trying to make the world a better place it's really been a good partnership yeah I'd echo that and that's just the rate of innovation in AI changes monthly so if you're going to be a good partner to your customers you've got to change just as fast so our partnership has been great in that space awesome next time we're out of time but next time come back and talk to a customer really want to understand and kind of dig into some of the great things that they're extracting from you guys so Charlie, Brian, thank you for joining Dave and me on theCUBE this afternoon thanks, thanks, thanks for Dave Vellante I'm Lisa Martin you're watching theCUBE, y'all from Purexillary in Austin, Texas