 From Austin, Texas, it's theCUBE, covering Pure Storage Accelerate 2019, brought to you by Pure Storage. Welcome back to theCUBE, the leader in live tech coverage. I'm Lisa Martin with Dave Vellante. We've got a pretty cool guest coming up next, guys. You may have seen him here on theCUBE before. He is back, Matt Harris, the head of IT for Mercedes AMG Petronas Motorsport. Matt, welcome back. Good afternoon, how are you? Good, we got the car over there, lots of excitement. One of the coolest sports I've ever become involved with, Formula One is this incredible mix of technology, strategy, all these crazy things. You guys at Mercedes have been partners, customers of Pure for about, what, four or five years now? 2015 as a customer, and we became partners in 2016. I wonder if they like to say Mercedes AMG Petronas Motorsport has had five consecutive years of both constructors championships, drivers championships. You're at a great position on both for 2019. Give us a little bit of a history about the product that you put out on track every other week and how Pure Storage is a facilitator of that. Yeah, okay, so it's an interesting story for those that are interested in Formula One, because what you see on the track looks the same, but realistically, every time it goes out the garage, it'll be different. That level of difference could be a simple wing change or configuration, always based on data that we're learning from during a race weekend. But every week, we also have a different car dependent on the track we're going to. So we have two different worlds that are basically we iterate on a minute by minute, hour by hour, and day by day at the track. But in the factory, that could be the same sort of iteration, but it could also be into weekly or monthly or year for a car. So all of that is based on data. So everything we do as a business is revolves around data. We never make a change to the car without being able to back it up with empirical knowledge. Even if the driver turns around and tells us they feel something called they believe something, we will always make sure we have data to back up that decision. So access to data is critical. Compute performance, whether it's high performance compute for our CFD, for instance, whether it's for you as an end user, access to data is critical across everything that we do. It's time critical. Time is our currency really as a business. If we slow down your job, generally that probably means that you've got less time to make the correct decision, or maybe you have to turn it into a guess or a hunch, which that's never a good place to be in our sport. No, I would thank God. If I recall from our conversation last year, there are rules that say how many people you can have in your entourage, like 60 I think it was, right? Yes. And at the time I think you said you got like 15 allocated to data. Is that ratio kind of still holding? Still exactly the same in our track side environment, that's still the same. In the factory, we have more than that. Depending on how many people and what time of day and what day of the week. So on a Friday, a practice day, we can have a minimum, there'll be 30 people in our race support room. We'll be looking at data along with those other 15. But you can have the whole aero department or design department or logistics, whoever can still be looking at data from the track real time. So we can have as many as 4 to 500 looking at data if they want to and if it's the right thing going on. Earlier in the season, you generally get more people looking as the season goes on. It's probably more aero focused, maybe mechanically if we've got something new or maybe the engine division. Again, in a completely separate building in the UK, 40 miles apart, they've got another set of people that will be looking and trawling through data real time from the, but looking really at the power unit rather than the chassis side. And you're generating like roughly a half a terabyte a weekend on a race weekend. Is that still about the same or is that growing? From a car perspective, it's just under half a terabyte, but we produce up to another half a terabyte of other supporting data, whether that GPS data, weather data, video, audio, whatever it would be, other information to help with the strategy side of things. So we're around seven, I would have said 750 to one terabyte per race weekend. And each car has about 300 sensors. I think when we spoke with you last year, or maybe a year and a half ago, it was about 200. So that's increasing in terms of all of the data being captured every race weekend. But one of the things that Madda says is, you know, where IT at Mercedes is not that unlike IT at other groups who really rely on high performance systems, but you do put out a new product every two weeks and this really extreme range of conditions, your product is extremely expensive and it's pretty sexy. But the portability factor, you have to set up IT shop, how many, 20 weekends a year, and set it up in what, 36 hours and take it down in six? Yeah, and I always joke about the taking it down in six is a bit like a Benny Hill sketch. It's obviously choreographed and well rehearsed, but we have all the same systems as any normal business would have. The track side environment is very different though. We don't have air conditioning, so all the IT equipment has to work at the natural ambient air temperature of the country we're in. This year, believe it or not, Germany and Hungary have been our biggest challenge we've had for the last probably three to four years, because they had 45 degree ambient air temperature. So forget humidity for a minute, which is another kettle of fish probably affects us a bit more maybe than the systems. But we're only chucking that air as fast as we can across the components. So we're not putting any cooling into what is probably around the tolerance of most IT systems. So we had to rely basically on air throughput to turn around and keep it cool. Now the benefit with Pure is actually it doesn't create any heat either. There's no real heat generation. So it's quite tolerant, which helps us because it doesn't create more. But the environment we put it into is quite special. But what we're doing is what any business would want to do, access to email, file systems, what we're trying to do is give it in a performance fashion. People need to make a decision. So in qualify, in for instance, those 300 sensors, that information that we've got from the car, we've got minutes to make a decision based on data. If it takes you too long to get the data off, you can't then look at the data to make a decision. So we have to make sure data ingest from the car and then basically multi-access from everybody in the factory or the track side is performant enough to make a decision before the car goes back out again. Otherwise, we're wasting track time. So you've always had data in this business, like early days was all analog, and it obviously progressed. In thinking about what you want to do going forward with data, what kind of information or capabilities don't you have that technology in the future could address? Yeah, so interesting one is technology in the future. If you know what it is, let me know. But with what we know right now, I think a lot of it's going to be about having the ability to have persistent storage, but actually the dynamic of the compute resource. So looking at things like Kubernetes or anything like that to turn around and have dynamic resource spin up as and when required to do high performance compute calculations based on the data, maybe to start giving us some automated information. I'm going to be careful with the ML AI. It's for our business. It's not quite as simple as others because our senior management are very technically capable and they just see it as advanced statistical analysis. So unless you program it, it's not going to give you an answer. Now, we've started to see some things this year where actually the computer is teaching us things we didn't ask it to. So we have got some areas where we're beginning to learn that that's not necessarily the case now, but for us, that access to data moving forward, it's probably going to be the compute combined with that underlying storage platform that's going to be critical. On stage you heard Rob and people talking about the ability to have that always present storage layer with the right compute. That's something for us is going to be critical because otherwise we're going to waste money and have resource sat doing nothing. Is security an issue for you? I mean, it's an issue for everybody, but is it a game of honor? Because you got this, you know, a little community that are you guys trying to hack each other's systems? So it's an interesting one inside the sport, actually no because a few years ago there was a very high profile case where data went between two teams and there was 100 million pound fines exclusion from the sport for a season. So that's too big. You don't mess with that, right? But also if you think of that from our perspective, we've got the Daimler star on here. We cannot afford to have any of that brand reputational rubbing off on Daimler. So that's a no-no. Other teams I can't talk for, but we're all fairly sensible between ourselves. What will be interesting moving forward is what technologies are in our sport that actually other, whether they're motor manufacturers or not is their technology in there that they're interested in? Maybe the battery technology from the power unit side of things. Is it the power unit itself? Are other things actually more interesting to those other places? Is it legal for you, by the rules of the sport, to monitor just data or capture data, whether it's visual or whatever, from your competitors? Yeah, so anything that's public is fair game. Okay. So we get given all the teams actually, we get a standard set of three or four different streams of information around GPS timing and some video feeds and audio feeds and they're publicly consumable by the teams when I say public for a second. And those feeds, we can do what we like with. You know, they're there for us to infer information, which we do a lot of, is what helps our strategy team to turn around and actually predict what we might or might not need to do as far as a pit stop or tire degradation. And that's where the human element must come in too, understanding the competition, like two football coaches who know each other, right? Well, yes. And now if you think, if you add to that, the human element of, well, what happens if one team's strategy person changes? Are they going to make a different call based on the same data? Is their hunch different? Do they think they know better? Yeah, within a team, you can have that discussion. So what happens in another team where their car's not as performance or their mindset maybe they're thinking differently? Or maybe a team's got the most performant car at the moment and they think that they're going to do X and we're like, well, we're going to do something different then to try and actually catch them out. So do we now do the normal thing? So let's look at- Great gamification, I love it. Exactly. Let's look at, I'll make a prediction. 2019 is going to be another Mercedes-AMG Petrinus Motorsport year. Shhh. Not good, there you go. Not good, there we go. So at the end of the season, all of the data that you have collected from the cars, all the sensors, all the weather data, GPS, et cetera, how does peer facilitate in the off season the design of the 2020 car, for example? Where does it play in things like computational fluid dynamics? Yeah, okay, so all of our production data is on pure. Whether it's on a rail or blade somewhere, it's on pure storage across the site. So they're involved whether you're talking about design, whether you're talking about finite element analysis for high performance, or the CFD, using high performance compute systems, everything's on pure. So from that point of view, it's making sure we're using the right resource in the right place to get the best performance. Now, CFD's an interesting one, because we're regulated by the FIA about the amount of compute that we can use. Now, because of that, you want it to be as efficient as you possibly can. It's not speed, but the efficient use of CPU time. So if a CPU is waiting for data, that's wasted, okay? So for us, it's trying to make sure that whole ecosystem is as efficient as we can. That's obviously an integral part of everything we do. So whether we're wind tunnel testing, whether we're in the dyno, the simulators, but everything basically comes back to trying to understand and correlate the six or seven different places we generate data, trying to make sure that when there's a change in the simulator, we understand that change in the real world or in a dyno or in CFD. So all of that, what pure do is allow us to have that single place to go and look, high performance, always available. And for me, I don't have to have a storage admin. You know, we've got a team of people that actually are thinking about that for us at Pure. You know, they're as invested in us these days. You know, I walk around here, I'm very fortunate. I get to see all of the senior guys here and they're asking me what's going on and house things with SQL or Oracle because they know exactly what we're doing and they're trying to say what's coming. So things like object engine and PSO, we've been talking to Pure about using that over the coming months, but what we're not having to do at the moment is go out and learn it. Actually, they're coming in and they're telling us all about it. So they become a virtual extension to my team, which is just amazing. Yeah, far more efficient. You're able to focus on a much more things that drive value for the business. As we look at some of the things like the Evergreen business model, what were some of the big, aha, Pure is the right solution for us back in 2015? Was that part of it? So Evergreen and Love Your Storage were two things at the time that were just incredible for us. Because Love Your Storage was basically, you could have an array and basically you could use it and there was no commitment, no anything, but if you liked it, you could keep it, obviously paying for it. And when we did that in the factory, basically within a week I've been in there, the team were like, whoa, hang on, that's going nowhere. So that was a nice, easy one. But Evergreen was an interesting one, which has only really, truly for me, I've always bought into it, but the last probably 18 months, we've used it time and time and time again because the improvements with the speed of X90 coming in, VME drives, when we were looking at capacity, what we did was we turned around and said, well, actually we can buy more dense units in an X90. So we're only buying the extra capacity but we were getting new technology. Right, innovations. All that innovation that Pure are putting into their products, we're getting it. So today when they were talking about the memory-based access and a few other things, I was sat there going, oh, I can use that. Oh, and there's no work for me either, there's no effort. The only thing I've got to worry about is whether I've got capacity for those modules to go in. So Evergreen has worked several times because I don't have to go back to the CAP export and go, can I have another X million pounds please? Why? I need some more storage. Yeah, but you bought some the other day. Yeah, well, that one, I'd need to get rid of it because I need a bigger one. And I don't have to do that now. I just go and I'm telling them what the increase is for, which actually they can choose then if they want the increase. They know what the business benefit is rather than just IT has got to turn around and either replace it because of age or the new version doesn't support, it's not an upgrade from the old one. I was looking at some of your stats in the case study that's currently online and I imagine these numbers have gone up. It's 68% reduction in data center rack space and saving 100,000 pounds a year in operating costs. Yeah, that would have been probably two years ago-ish roughly those figures. And the operating cost is a huge improvement for us. CAPEX is probably the biggest one for me though and moving forward with cost caps coming into Formula One, that type of thing is going to be invaluable because not having to do a forklift upgrade of your storage. Well, I wouldn't know what I would do if I had to upgrade what I'd now own from Pure. I can't even imagine what, I don't want to turn around and tell my bosses what that's going to cost. Well, it sounds like you really attacked the OPEX side with R&D, with Pure R&D. I kind of like that, shifting labor to R&D because you don't want to spend labor on managing storage arrays. It makes no sense for your business. Okay, what do you want Pure to spend R&D on now? What problem can they solve for you? Where's the pain? So, Array C is going to help, if I'm really honest. That actually is going to help fill a hole quite well for us because we weren't really sure what to put some of that less hot data. And we were like, well, where are we going to start to put this now? Because we were beginning to fill up the array and the blades. But actually with Array C now, we can actually use that different class of storage actually to keep it still online, still be able to do some machine learning AI in the future when that comes around. But actually I can now have more longevity out of my existing array and blades. So that's brilliant and coming. I think having, I need to be careful because I know some things that are coming. The active sync in Array is brilliant and we've been using that since it came out. Having that similar or same ability in blade when it comes will be very advantageous. Yeah, having those blade enclosures, we've gone to multi chassis flash blade over the last six weeks. So that for us is great. Once we can start to synchronize between those two then that's another big one for us for resiliency, for fault tolerance, but also workload movement. That thing I said about a persistent storage layer, I'm not going to need to care where it is and it will be worked out by the storage and that orchestration layer. So it can have the storage and the compute in the right place. Wow, great story Matt as always. And I think as Pierre calls this the unfair advantage coming to life best of luck for the rest of the 2019 season. I'll take it. Thank you. All right, we'll see you next time. Thank you for it. For Dave Vellante, I'm Lisa Martin. You're watching theCUBE from Pure Accelerate in Austin, Texas.