 Live from the BuildGram Auditorium in San Francisco, it's theCUBE, covering Pure Storage Accelerate 2018. Brought to you by Pure Storage. Welcome back to theCUBE. We are live at Pure Storage Accelerate 2018. We are in San Francisco at the BuildGram Civic Auditorium. This is a really cool building built in 1915. Loads of history with artists. I'm with Dave Vellante. I'm wearing prints today in honor of the venue and we're excited to be joined by a longtime Pure Storage customer, Mercedes-AMG Petronas Motorsport, head of IT, Matt Harris. Matt, it's great to see you again. Hey, good morning, I should say. I think it is still morning somewhere. So Matt, for folks who aren't that familiar with Formula One, one of the things you know I'm a fan, it's such a data-intense sport. You've got to set up a data center 21 times a year across the globe with dramatically different weather conditions, humidity, et cetera. Give our viewers an idea of your role as head of IT and what it is that your team needs to enable the drivers to do. Okay, so in general terms, we're like any other normal business around the world. Yeah, we have huge amounts of data created depending on what your company is doing. Ours comes from two cars going around the track. That is the lifeblood of our work, our day work and all that data is always analyzed to work out how we can improve the car. But what we really have is an infrastructure, the same as many other companies. We have some slight differences, as you say. We go to 21 countries. In those countries we turn around and we have 36 hours roughly to put everything together in a different world, different place and then everybody turns up and uses it as though it's a branch office. 100 people roughly sat there working in the normal environment. We use it for five days and then we take it apart in six hours, put it into boxes, take it to another country. And we do the same thing again. We do that 21 times, sometimes back to back, sometimes with a week in between. Week in between's quite easy. Back to back sometimes. We go from Canada, maybe all the way across the world from Monaco within the space of a week. So we've got the flights in the way and everything else. And we also end up having an engineer car, run a car around the track and hopefully win races. So you basically got a data kit that you take around with you. And what did you do before you had this capability? Was it just gut feel? Was it finger in the wind? For about 15 years, we've been running all everybody's classes and internet of things. We've been doing for about 15, 20 years, the car. It's got around these days around 300 sensors on it. Without those sensors, realistically, we'd be running the car blind and we probably couldn't even start the car, let alone actually run it these days or improve things. We turn around and we're always ingesting data from the cars, real time, that real time data, actually we transfer to the garage. That's no problem at all, but we also bring it back to the factory because we're limited on the number of people that are allowed to travel with the team. So we're physically only allowed to take 60 people rules to tell us we can only take 60 people to work on the car. Now of those around about 15 are probably looking at data. We're generating around about half a terabyte per race we can these days. In 15 people, it's not enough eyes realistically to turn around and look at all that data all the time. So we take it back to the UK and in the UK, again, we have anywhere between another 30 and maybe 800 staff will be looking at that data to help analyze particularly on a Friday. Friday's about running the car and learning. We discussed a few minutes ago, what's the weather like? What are the tires like? What's the track like? Has there been any change in track? Has it been resurfaced? What's going on with the car compared to what we think is it's optimum? And on a Friday's, iterative change and learning about tire degradation, tire life, tire wear, the weather conditions, how they're going to interact with the car, all based on data. The interesting thing for me has always been that we have all this data, but the two drivers in the car are the biggest sensor for us. They turn around and tell us how they felt. When they were going round corners, was it good, bad, indifferent? But as soon as they tell us something, we always go to data. We've taken their interpretation of how their body felt. We turn around and then look at the data to prove what they've told us. So, an interesting anecdote very quickly. Last year, in Singapore, Valterri was going across the bridge, and he said he could feel that the throttle felt like it was cutting, and we couldn't see it in data, and we were looking and looking, and eventually he said, no, it absolutely happens every time I cross the bridge. And they found a 20 millisecond gap in throttle application, basically because there was a magnetic field that the bridge was creating, so a sensor was actually cutting the throttle. He could feel it, we could eventually see it in data, shielded the sensor, everybody's happy. So, you go from the human being could feel a 20 millisecond gap in throttle application for us finding it in data, engineering a solution and changing things. So the human's still a critical part of the equation. So where does pure storage fit into this whole thing and give us the before and after on that? So, three years ago, we started working with pure because I have two different solutions, one of the track and one in the factory. One of the track, realistically, I have some constraints around space, power, heat. The most people would love to take the racks as we were talking about, we take around the world, they would love to leave a nice air conditioned off it, computer room in a, and just leave it there all year. We move it around, but that rack of information, we have to spend $298 per kilo to transport IT equipment around, well, any equipment around the world. So that's, we've got tons of equipment that we take around the world, it's thousands and thousands of pounds of freight cost. So we went from 40U of old school spinning disk, lots of complexity in cabling, administration, down to two, three U M20 arrays. Now they're more heat tolerant. I have two power cables in each and two network cables. So complexity is gone. It just works, it's heat tolerant, it doesn't create a lot of heat, so I haven't got the added issue of that. It's not using a huge amount of power, so my UPS solution has to be smaller. So everything just got smaller, cheaper. Really simply at the track, we improved the performance for everybody. From an IT point of view, we got very, very simple, incredibly easy to look after and manage, but it's very reliable and performant at the same time. We then went to the factory, where I've got 800 people looking at data. The problem is when a car goes round and we offload it, there's one single file. We haven't got this distributed amount of data that everybody, so you've got one file that everybody's trying to open. Old school disk, you've now got contention for that one file that everybody's opening. So people would come back from the track and go, why is it so slow to open information in the factory compared to at the track? Trying to explain to them contention of data in those days was a little bit difficult, but now we have 800 people that don't need to care, and why that matters for us is decision-making. So if you think about qualifying, those that don't understand Formula One, we have three sessions of qualifying, and the car goes out roughly two times in each qualifying session with around about a couple of minute gap in between the times the car goes out. That couple of minutes is about changing the car to be optimal for the next run. If it takes you minutes and minutes to offload data, open the data, review the information that the driver told you and make a change, you can't go back out a second time. So everything's about optimal performance for those engineers to optimize the performance of the car. What we're able to do now is to turn around and make sure that we're making the correct decisions because rather than data taking two or three minutes to open, it's in seconds instead. So you can look at the data, make an informed decision, change the car, hopefully improve every time the car goes out. One of the things, Matt, that Charlie Giancarlo, the CEO of Peer Storage said this morning during the keynote was that less than half a percent of data in the world is analyzed. Talk to us about what Peer Storage is able to facilitate for your team to be able to analyze that data. How much of that data are you able to analyze and talk to us about the speed, criticality? Yeah, okay, so quite a lot of the work over the previous, probably 10 or 15 years has been very human-centric. So it's what data I know I need to go and look at to understand to be able to compute, to turn around and maybe infer information from to be able to make a better decision. So strategy is probably one of the best places these days where the data that we're learning all the time. We have data about ourselves, but we also have data about the other teams. Those teams have the same data about us as well. You know, GPS data, timing data so we know what's going on so we can infer information on a competitor as well as ourselves. Tire degradation, tire wear, tire life, all things that you can infer that mean that you were mentioned earlier on about a pit stop. If a safety car comes out, should you pit? Shouldn't you pit? Those decisions are now based on accurate data about whether we think a competitor will pit, whether we think the competitor's tires will last. Can we overtake that competitor because actually the track does or doesn't allow overtaking? So lots of decisions made real time based on exactly what's happening now, but inferred from previous races and we're always learning all the time. Everything is about the previous races, information we're learning every time. And how much of that heavy lifting of that data is machines versus humans? Are the machines increasingly, I don't want to say making the decisions, but helping make the decisions? Yeah, so we're not in a position at the moment where the machines are making decisions. They're helping us to be informed, to visualize. Yeah, we work with the likes of Tibco as well as Pure and other partners and sponsors that we have where they turn around and actually they help us to visualize that data. The problem we've got at the moment is we're still looking at all the data where we really want to get to is looking at exceptions. So actually the norm, don't show us that data. We don't need to know, don't need to care. Not the outliers. We want the outliers that our problem though is that our car changes every time it goes out. So an outlier could be because we've made a change. So now you've got to still have some human that's helping at the moment. So we're trying to understand how we can use machine learning techniques. In certain places we can, so image recognition and other bits and pieces like that, we can actually start to take advantage of. But decisions necessarily around configuration and the next change to the car at the moment is still indicators given to us by simulation and then a human at the end of the day is making the decision. And the data that you talked about that is on your competitors, is that a shared data source or is that, but it is. Everybody shares the same data. Every car has a transponder on it. Basically gives GPS with longitude, latitude and all sorts but incredibly accurate. If you consider the cars are doing 200 mile an hour, we have an accuracy of around about, it's less than 10 centimeters accuracy at 200 miles per hour. Now, if you think of your GPS on your phone, you struggle to know whether you're on the right street sometimes. But your differentiation there is your speed at which you can analyze the data, your algorithms, your skill sets, your talent. And then obviously we're here at Pure. There's a component of that speed which is Pure. Aren't you worried that your competitors are going to get your secrets or is everybody in the track use Pure Storage? Everybody is turning around and using their own methodologies, their main, their own software. The thing for us at the moment is to make sure that we keep the really secret things ourselves, our IP sensitive, keep those to ourselves. So what we do with our storage, people know about and other teams are copying and seeing the advantages of Pure as well as some of the other tools and partners we partner with. The benefit of us though is that we have a partnership with Pure not just to purchase in. So we've known about some of the products, so FlashBlade, we know about a long time before it is released. Yeah, we work with the team on what's coming, we know some of the advances in the technology before it's live. And that's critical for us because we can get a stitch, a march on everybody else. Even if we're six months ahead of somebody else on a technology or a way of doing something, six months is a long time in F1. So, sorry to say, Pure calls us the unfair advantage. And you are, Mercedes has last fall won the fourth consecutive constructors championship coincidence, I don't know. But talk to us about this sort of symbiotic relationship. Are you also able to help influence the design of the technologies at Pure? Yeah, so I wouldn't say that we help design necessarily, but they'll take into consideration our requirements and I wish it's like a number of other people that will be here. You've heard other people talking on stage and we'll always be talking about what we would like to be doing. What we could be doing if we had some new technology, whether it's S3 connectivity to the flashblades, whether it's NFS, whether it's SIF, whatever that would be, the containerization of the storage front end. Whatever that would be, we're always talking about how we can work with the Pure storage to improve what we're doing. So ideally, IT, get out the way of the business. My ideal is that IT's not seen, it's not heard and it just works. Obviously in IT, that's not always the case, but. I want to unpack something that you said earlier. You said, I believe two or three years ago, three years ago that you brought in Pure and you had substantial performance improvements. I talked to a lot of customers and what they'll typically do in that situation is they'll compare what they saw in 2015 with what they replaced, which was probably a five or eight year old array. True in your case or not, if it is true, which I suspect it is, it had to be something else that led you to Pure because you could have bought the incumbents all flash array and got much better performance, but first of all, true or not and what was it that led you to Pure to switch from the incumbent, which is not trivial. So quickly, was it five or eight year old hardware? In some places, yes, some places no. So it wasn't, we took a decision to take a step back and look at storage from a different standpoint because we just kept adding more disk to try and get around an issue. And we've got a fairly strange data model to compute. We don't need much compute, we need lots of storage. So some of the models that were talked about on stage where I need, Matt Burr was talking about the fact of I want some more storage, you need to buy some compute. And that was just so annoying for us. So there was different reasons, but the end goal, you're quite right, performance. Yeah, we could have got it probably from anywhere. Being brutally honest, lots of other technologies could give the performance because we don't give that level of performance maybe of a service now or a big financial institution. We've got data, it's important. We've got critical time scales to open and save data. Okay, critical to us as far as erasing. But what was important for me was simplicity. Absolutely. Now we got other benefits. The evergreen model was brilliant for us, but simplicity was critical. We had a storage guy that was spending his life managing storage. Nobody manages storage now. They turn them in and they go into VMware, they want a new VMware server. They just spin it up and the disk is associated. We don't have to think about it. You don't have that storage specialist any longer. You know, we started working with other partners, you know, Rubrik, for instance, integration with the pure arrays as well. Again, enabling us to get out the way and not having to worry about backup. Traditionally, we'd headed a guy that was always changing tape. I saw on the slides several times today about tape archive. I'm going, I never want to see a tape archive. I just don't care about it any longer. I just want to be able to turn around and give the business the SLAs they want on their data and then not care about it. Also, can I then still turn around and mine that data in those archive or backup, not backup, but in the archive location? So there's huge differences, but simple is the best thing for me. We have a small IT team that we have to look after a huge amount of kit. And if it's complex, it's just, I can't employ the right people. Simplicity, performance, portability. You mentioned integration. You've got a big partner ecosystem here that, so they're having the ability to integrate seamlessly with Rubrik, Tipco, et cetera, is key. And yeah, for us, the partners are an extension of the team, my team in particular, because I can't turn around and just keep adding staff. We have to look after the day-to-day and keep the lights on. But I can't just keep adding staff to look after a new technology. It needs to look after itself. So the simplicity is absolutely. Performance was a sort of a no-brainer. Evergreen was a brilliant one for us because just not having to do those forklift upgrades, I think in the three years, we've gone from M450s to M70s. We've gone from M20s to M50s, M50R2s. We've done all of these. I've been stood on stage before on a day when we've been doing an upgrade during the time I've been stood on stage. Yeah, and so people talk about the forklift upgrade. I don't have to worry about it. It doesn't happen. Totally non-disruptive. Yeah, yeah. And what, you do change out the controllers, right? Yeah, so we've changed out controllers. We've done all sorts. We've gone from capacity upgrades, so complete shells of discs, completely different, gone from, I can't remember the exact size, from two terabyte to three terabyte drives, new controllers to give us the new functionality with the NVMe. All during the day, we don't do it out of hours. There's a lot of the business, a scared stiff, when we turn around and they go, oh no, no, no, but we're running the wind tunnel, we're doing this CFD, we're going, it doesn't matter. Zero downtime. It doesn't matter, zero. No planned, obviously no unplanned downtime. It is planned for us, but no planned downtime. But the user doesn't see it. No, no performance, no downtime, no nothing. That's nirvana for IT. Yeah, well it means I don't have to keep asking people to do long shifts through the night to do a simple upgrade. What should be a simple upgrade? Get your weekends and nights back. Hopefully. We end up racing those, unfortunately. Okay, but that's the fun stuff. Yeah, yeah. That is the fun stuff. And for those who aren't that familiar with Formula One, I encourage you to check it out. It's one of the coolest strategic sports that is really fueled by technology. It's amazing. Without technology, honestly the cars wouldn't be anywhere near what they are today. IT systems, we underpin everything that the company does. Nobody really wants to say that IT's the lifeblood of a company, they don't. But we need to be able to deliver and actually let the business actually take on new technologies, new techniques and get out of the way. So we've got a huge amount of work. A lot of what Charlie said on stage earlier on, I've been having conversations with the guys here about autonomous data centers, immutable infrastructure. It's critical for us to get out of the way and allow business to, if they want some new VMs, new storage, it just happens. Not need a person to be in the way. Wow, make it sound so simple. Well one of your primary sensors, Lewis Hamilton is currently in the number one position, the Altree Bot Test in third. Monaco coming up this weekend, introduction of a new Hypersoft tire, some pretty exciting stuff. Yeah, so the Hypersoft tire is going to be interesting. First race with it. Is it before the Monaco track? Yeah, so they originally designed it for Monaco. I believe it'll go to another race as well in the short term, but we didn't even run it in winter testing earlier in the year. So the first time we ran it was actually Barcelona test last week. I've actually heard nothing about it, so I don't know whether it's good, bad or indifferent. I don't know what's going to happen. But it's going to be an interesting week because it's a very different track to where we've been to so far. Traditionally, some of the other teams are quite strong there. So this weekend's going to be an interesting one to see where we end up. Monaco is always exciting, Grace. Matt, thanks so much for stopping by the Cube and sharing with us what you're doing and how you're enabling technology to drive this board. It's really cool. You're welcome, no problem at all. Again, I'm Lisa Martin with Dave Vellante, live at Pure Storage Accelerate 2018. We're at the Bill Graham Civic. I'm Prince for the day. Stick around, Dave and I will be right back with our next guest.