Latest / Technology Now / How Mercedes-AMG PETRONAS F1 Team uses 7 billion data points per race | Michael Taylor
Transcript
- MICHAEL TAYLOR what many don't realize is, the race on track is just one part of the competitive nature of Formula One. Although we're classed as a sport, when you think of the car, it's effectively a rolling laboratory that changes from race to race, even sometimes from session to session, during a Grand Prix weekend. How we're able to do that is underpinned by technology. Everything that every single person does and contributes to the organization, we talk about contributing our own 10th of a second to the performance of the team, SAM JARRELL Wow. Well, I mean, given all of the data that they have to process, both when they're practicing racing, and then during the actual race itself, just to save, like, 0.0001% of a second or something like that, it doesn't surprise me that they sort of view this... view the F1 races as, like, a high-speed lab. Though that seems kind of scary to me, if I'm honest. MICHAEL BIRD Yeah. The thing I love about the sport is just they are looking for time everywhere. How can they save time? How can they get around the track just, like, a thousandth of a second faster? Which is basically what we're gonna be talking about today, because we're gonna be exploring the technology behind making a race-winning car. I'm Michael Bird. SAM JARRELL I'm Sam Jarrell And welcome to Technology Now from HPE MICHAEL BIRD Formula 1 isn't just about the driving. The amount going on behind the scenes to scrape every last millisecond off the time it takes to get around a track is absolutely monumental. SAM JARRELL Right. I mean, you only need to see the speed of a pit stop to be blown away by the efficiency of the team keeping the cars running. Two and a half seconds to change four tires is pretty mind-boggling to me..i MICHAEL BIRD Yeah, last time I had the tires changed on my car, I think it took about four hours. So, yeah- ... the fact that they do it in about 2.1 seconds is pretty incredible. And it goes even further than that, too, because off the track, you've also got all of the design team who are following the performance of different parts of the car, ready to make changes before the next race, as well as all of the IT team who are analyzing race data, and this includes our guest for today's episode. So Michael Taylor is the IT director for the Mercedes-AMG Petronas Formula 1 team. And as you'll see, because this is a video episode, so be sure to check out wherever you watch your podcasts, we did this interview standing next to their new car, as well as a trophy from one of their recent wins. SAM JARRELL Can you imagine rolling up to the office in that car? MICHAEL BIRD I think it'd be quite fun to drive on, like, a, like, a highway or a motorway. I think that'd be really fun. You'd definitely be the first to work, wouldn't you? SAM JARRELL I think that, uh, you would be the first to work, but you might have, like, a trail of police cars- ... very, very far back trying to follow you. But that's enough about us getting excited being up close and personal with a real-life F1 car. What did you and Michael talk about? MICHAEL BIRD Well, Sam, I wanted to start off with the basics. So the first thing Michael gave me was an overview of the role that technology plays in an F1 team. MICHAEL TAYLOR First of all, Michael, thanks very much for having me. It's a pleasure to be here. gosh, technologies in Formula 1, where do we begin and where do we end? we use pretty much everything that's out there, as you can imagine. Obviously, the car is a significant focus, capturing time series data. So there's a huge amount of, infrastructure, networking, compute, hypervisor required to deliver those, services, certainly track side. And then back in the factory, we do actually refer to it as a factory, so we still have a working factory. Half the facility is producing, both metallic and composite components, so there's a huge amount of, IoT-based technology in there, and ultimately, we also have a business to run, so we have all of the traditional functions like finance, HR, marketing, and legal, and we have all of the, the technologies and systems that every other organization across the world has to underpin those services. MICHAEL BIRD and how has that changed in the last, I don't know, like the first season I watched from start to finish was 2009, like how has the technology changed since then? Because IT was in a completely different place back in what, two decades ago nearly. MICHAEL TAYLOR Yes, it was. the evolution of technology through the sport has pretty much followed, the general wider world. so the ECU within the car has been through a couple of major significant evolutions, which obviously each time brings more capability, more functionality, and the ability to capture more data. But then just generally across the organization, data has always been the lifeblood of a Formula One team. historically though, that was focused in the performance areas, primarily the car and then how we support the design and development of the car. These days, with compute and storage so prevalent, data is everywhere, and we instrument absolutely everything we can across our organization, always looking for how we can find that next millisecond. So we measure our performance in milliseconds, and obviously the key thing around data is what's the next decision that it decides for you. MICHAEL BIRD Yeah, and I've heard stories from different teams around the grid of, it's not necessarily just the car where you find milliseconds. MICHAEL TAYLOR you're absolutely right, and what many don't realize is, the race on track is just one part of the competitive nature of Formula One. we look at every aspect of our organization in every single possible way, constantly analyzing how we can improve things, how we can learn, how we can be better. MICHAEL BIRD and technology's a part of that. I mean, it's presumably, the lifeblood of the organization. So performance is there in the technology as well? MICHAEL TAYLOR Technology is, is an enabler. Although we're classed as a sport, when you think of the car, it's effectively a rolling laboratory that changes from race to race, even sometimes from session to session, during a Grand Prix weekend. How we're able to do that is underpinned by technology. Everything that every single person does and contributes to the organization, we talk about contributing our own 10th of a second to the performance of the team, technology underpins in some way, shape, or form. MICHAEL BIRD Okay. So let's talk about the car really briefly because, you mentioned that this car is packed full of sensors. It is indeed, yeah. MICHAEL TAYLOR So we'll start the weekend with probably in the region of about 300 sensors on the car. That will be generating multiple thousands of channels of data. we see both in the back of the garage at the track, but also streaming as near to real time as possible back to the operations center back in Brackley and also Brixworth. the sensors are dotted all around the car, for two reasons. One is to ensure the car is performing as expected, and obviously there's very fine tolerances and fine margins So making sure that things are working as they should do, as per design, we have sensor data to, to validate and back that up. But then equally, the secondary purpose of it is really to understand the performance profile of the car. So using the sensor data as part of the validation path of feedback from the drivers when they say they're not happy with how the car feels in terms of corner entry or stability, or there's, too much understeer, too much oversteer. And that sensor data then is used to, evolve the setup and configuration of the car over the course of the weekend, and then obviously feeds into the learning loop then for the next event. Now, in terms of how much data we're generating, it's a significant amount. a race will generate upwards of seven billion data points, so a single car during a race will generate upwards of seven billion data points. MICHAEL BIRD upwards of seven billion data points. Uh, how, and how often is one sensor sending data back? MICHAEL TAYLOR Uh, it very much varies on the configuration. Sometimes you'll have, uh, one sample a second, sometimes you'll have 100 samples of a, a second. It just very much depends on, on the sensor. Now, time series data, we, we have two types of data that come off the car. The real time data stream, that's limited unfortunately by bandwidth. Uh, we also have a data logger on the car, so within the ECU, and we're able to generate a much higher fidelity there. So when the car comes back into, into the garage or the box, as it's called, we roll it back into the garage, plug it into the network like any other ethernet node, and we download the data from there and then fill in any of the blanks, because sometimes out on track, as you can imagine, we don't have perfect coverage of every single sensor. MICHAEL BIRD So there's like a little aerial on here that's sending data back to the, back to the, the, um, the garage, but it's not all the data because there's just not enough bandwidth. MICHAEL TAYLOR Exactly that, yeah. And they're all generally down on the front of the car. MICHAEL BIRD of the car. Wow, okay. And so yeah, at the end of the race, you plug the car in, you download all the data, and then you can make sense of what's going on. So I guess like in a testing session, if you're trying different wings or different profiles, you'll try that, download the data and go, "Yeah, that didn't work. The driver wasn't happy, and actually we can see why he wasn't happy because of, you know, understeer, oversteer, whatever." MICHAEL TAYLOR So, so we do have probably enough fidelity from the real-time data stream to be able to see those things. But yes, we validate that then with, uh, the data that we download from the logger. We obviously do a lot of post-processing on that data as well, which is why it's really important to highlight the, the track side really is just the edge. And the car being a data generator, we then stream that data back to the factory in Brackley and Brixworth, or factories in Brackley and Brixworth as well. And we're doing some, uh, some additional post-processing on there, so enhancing the number of channels that we're capturing, um, in the process of, uh, of kind of evolving sensor technology so we can bring different data streams together to give us a much clearer picture of exactly what's going on, both from a reliability but also a performance profile. All obviously focused on how we find that next few milliseconds of performance for, for car and driver. MICHAEL BIRD So, um, I've been to, I've been to the factory in Brackley and, it's all done, through computers. Can you just talk through, like, how is a car designed? Presumably there's, um, CFD and wind tunnels and, and all that sort of stuff, MICHAEL TAYLOR Again, it's back to that technology underpins everything story. so generally, the surfaces that you see on the car, so the aerodynamic surfaces, they are the biggest generator of performance, so a mix of downforce and drag. the CFD process starts... So computational fluid dynamics, in very simple terms, it's effectively, the virtual modeling of wind passing across the surfaces of, different geometries or different areas of the car. We use that purely for experimental purposes, so that's where we do, the bulk of our experimentation and simulation and learning. So that's a really key part of the initial design process. So anything that we can do there in terms of, turnaround time of jobs, getting the, the results back to the engineers, maximizing the amount of compute we're able to use in a period. So aero testing is actually regulated, for all teams, and we're limited by, how much teraflop capacity we're able to use in a period, but equally, how much wind on we're allowed in the tunnel. So CFD is the starting point. andCFD will generate upwards of 50 terabytes of data a week. what looks to be working, we put that into physical form on a 60% scale model in the wind tunnel. Now, the wind tunnel, model, the spine effectively is full of sensors to make sure the, the gains seen in CFD actually are likely to deliver on track. Now, obviously, being a 60% scale model, there's still further calculation that needs to be done to make sure that is the case physically on track. But if it passes those two tests, it then gets handed over to the design teams No drawing boards there these days. All done very much digitally on our, design platform and data management platform, and those guys then have the responsibility to turn these beautiful shapes that are created into something that can be physically made. MICHAEL BIRD 'Cause, you don't stop developing a car. You don't, you don't build a car on the 1st of January and go, "That's it." Like, it's constantly changing, new parts added. So this is just an ongoing process. MICHAEL TAYLOR Yeah, absolutely. and every week the car is different in some subtle way, shape, or form. you're also very right that the car that starts the season, if we stop developing the car, by midpoint we'd likely be last. it's a constant battle back at the factory to arm the race team with the hardware that makes that car go faster to ensure that they can then execute a successful race weekend. MICHAEL BIRD So, from that point then, the technology that, you and your team provide to the team is vital for the development of the car. has that evolved over the last 10 years? MICHAEL TAYLOR Yeah, absolutely. the development curves have increased dramatically. Go back 10, 12 years, if you're able to deliver a second of performance throughout the course of a season, that was a really impressive achievement. These days, teams are aiming to significantly exceed on that. That's only been, made possible by just the amount of technology that we now have our hands on, technology is, definitely an enabler. Obviously, the data that we generate across all of the different experimentation, simulation, modeling, and then the physical car being on track, The advancements in software development, software engineering, data analysis, data science, you know, all of those things all contribute to the overall performance profile. MICHAEL BIRD Okay, can I talk networking? Because we've talked about grabbing this data off the car. How important is networking to the way the team operates, both in the factory and at the track? MICHAEL TAYLOR obviously we're very limited in terms of how much time we have on track with the car. So a traditional race weekend, we'll have two Friday practice sessions, one hour each, so two hours in total. We then have another hour on a Saturday morning before we go into qualifying. Qualifying is all about single lap performance. And then that's it until the race on a Sunday, so very limited time on track. So the importance of the network is, that it enables the team to just go about their tasks, their role, their responsibilities without a second thought. the car is on track, we need to make sure data is flowing off of the car, the engineers have visibility of that data, and they're able to act upon that data. for any reason that, that process falls over, you immediately, you're on the back foot then. So networking is key. It underpins everything that we do, so enabling those squiggly lines to make their way to the engineer screens, and then all the way back to, Brackley as well. added to that, we have huge, amount of hospitality and hosting and partners and marketing activities that we need to do at track. So it's not solely just about the racing part. We also have to enable all other services that allow the team, to perform, to the best of their abilities throughout the course of the weekend, MICHAEL BIRD how do you get that data back to Brackley, Do you just get a big internet pipe and you just plug in and jobs are good then, or is there- MICHAEL TAYLOR only it were that simple. carrying network traffic from locations that you only use for maybe six days a year is quite expensive. And obviously we operate under a cost cap, which means we're limited in terms of how much money we can spend on the design, development, creation, manufacturing, and racing of the car. Typically, IT falls within that cap. These days it's MPLS. so we, implemented, Edge Connect SD-WANs a couple of seasons ago. We're now able to supplement very expensive MPLS connectivity that's provided to us as a managed service across all 24 locations throughout the globe. with, direct internet access and other forms, say satellite type comms where appropriate, but even 4 and 5G type services too. and the beauty of that is that we're now in a much better place from a cost perspective, but also from a bandwidth perspective. We're able to carry more data back and forth. We're able to enable more services. As we start to expand a little more into the cloud, we're able to then start to divide traffic and, make sure that it gets to, the point that it needs to get to as quickly as possible. MICHAEL BIRD And sort of aggregating different connectivity types. Presumably, that also gives you a bit of, fail-safe. You know, if one goes down, you... Okay, they might be smaller, but you can still send some data back. MICHAEL TAYLOR Absolutely, absolutely 100%. Yeah. Availability and reliability, most important things, and then performance closely follows MICHAEL BIRD has that ever happened during a race where a feed has been cut? Has that happened in Formula 1 at all. Has that happened for you? MICHAEL TAYLOR it does happen, yeah. we've managed to find ways to, reduce the impact of those types of failures. So obviously we talk about reliability and availability- Yeah ... as really key metrics, and they are, because it needs to be up to enable everyone to do their job. some of the locations we go to, are very remote. They don't necessarily have the same kind of standards, qualities in terms of data centers and where, cables are being run so yeah, the, the SD-WAN has been a real enabler in terms of bringing that kind of flexibility, but also the reliability that's needed. MICHAEL BIRD Which leads me really nicely onto the sort of mobile data center. Presumably, you bring infrastructure with you to races. how does that look? How does that work? MICHAEL TAYLOR I'd love to say we have an A and a B rack that we send, to alternating locations around the world. But no, unfortunately we don't because of the cost gap. we take a, limited amount of infrastructure to the edge. It's vital that we can run these cars safely. if there were to be some kind of WAN outage, or connectivity issues, it's vital that we're able to run the cars and run the cars safely on track, and all teams obviously work under the same principles. So we take a relatively small, amount of compute to the track. You're right, environmentally it's a really difficult location. The racks are custom built, Still air-cooled these days, but we also use equipment that's validated to run at much higher temperatures. And we actually mirror across two racks, so if we do have any kind of failure or any issue, we're able to just fail over onto one rack and continue to maintain services. It's a fairly small, virtual environment there. We run probably just upwards of about 100 VMs, and there's about 150 terabytes of fully mirrored storage that's used. We then obviously have the connectivity that enables us to take data back and forth, from track to factory to be able to do more with it. MICHAEL BIRD Do you think you'd ever move towards water cooling? MICHAEL TAYLOR we'll definitely look at it, but logistically it adds a bit more complexity freight weight, so we have to fly all that equipment around the world. That's a significant cost- Yeah ... again, within the cap. So we always look at how we can take weight out of the racks, weight out of our infrastructure to ensure that, we're delivering exactly what we need, but at the minimum cost, available. MICHAEL BIRD which track or environment is, would you say the harshest environment for this equipment? MICHAEL TAYLOR Bahrain is notoriously difficult because of the time of year when we go to visit, they often have, sandstorms. And sand is so fine grain and floating around in the air, you can't necessarily see it. but if you look at the filters after a couple of days in Bahrain, you'll understand why. If any of that does make its way into, into the hardware, we can very quickly run into, issues. Equally, other locations, Singapore's always very hot and very humid. and some of the exhaust air that we're blowing off of the rack is pretty warm as well. So making sure that goes out of the garage and not heating up an already, hot engineering crew is vitally important too. But yeah, across the course of the calendar, there's many different challenges depending on geography, locale, local weather, all of those things. But they're just things we take in our stride in the world of Formula 1. MICHAEL BIRD so F1 regulations are incredibly strict. everything is, is, is an opportunity for performance, and so often that sometimes gets locked down. Um, my understanding is finance is also now part of the regulations. Has been for a few years. is that something that the technology helps with? MICHAEL TAYLOR the financial regulations were effectively the third tier of governance introduced to the sport in 2021. Prior to that, we had the sporting regulations and the technical regulations which were well known and fairly consistent throughout. The financial regulations have added an extra dimension of challenge to all teams. And technology spend, unfortunately IT sits squarely within the cap, so every technology spend we now make, there has to be some kind of justification, a genuine business case that, that adds up. I always talk about whatever we do investment wise needs to deliver value and move the needle. the financial regulations have evolved as well over the last few years, and we had another change at the start of this season which has added an additional level of complexity. Just as you start to think you're mastering the understanding of the regulations, they change and then you have different treatment for different things, and you then have to go and find ways to be smarter and cleverer again. obviously technology is a huge enabler to be able to do that large data analysis. we're now running as much data analysis against our financial forecast and the modeling as we are on some elements of the development of the, the car. And historically in Formula 1, it was all about, finding an opportunity to improve a part and taking that to the track as quickly as you possibly can and the financial aspects of that were sometimes a bit of a second thought. Now it needs to be part of the equation. So there's a cultural change that's happening certainly in our team, and I suspect all of the other teams up and down the grid to really, work out how best to evolve and how to manage another constraint in a world of, constrained regulations. MICHAEL BIRD Yeah. Michael, thank you so much for your time. It's been a real pleasure interviewing you on Technology Now. MICHAEL TAYLOR No, uh, Michael, it's been an absolute pleasure, too. Yeah, so MICHAEL BIRD Thank you very much. Thank you. SAM JARRELL Wow. Well, we often think, I think, of Formula 1 as, like, a battle between the drivers, but I think your conversation with Michael really highlights that it's equally a battle between information systems. I mean, the one phrase that kept coming up- Yeah ... is that, like, data is sort of the lifeblood of the team. Do you think that it's more of, like, the- a battle between technology, not just a battle between the drivers? MICHAEL BIRD Yeah. Yeah, definitely. And I think what I took away from the conversation was that they are looking for performance everywhere. It- it's not just about faster engines or front wings. It's just, it's every single element of how the team operates. SAM JARRELL Yeah, I agree. I think, uh, many of these ideas are transferable. I think most- Yeah ... organizations aren't trying to win a Formula 1 race, but many are trying to make faster, smarter decisions using data. And I mean, like, you guys had a point around how, like, development never stops the car. You don't just build it in January and that's your car for the year. But, like, the constant and continuous improvement, measuring performance, uncovering every tiny small gain is relevant whether you're building race cars or I think operating a supply chain or running a business. MICHAEL BIRD And adapting. I mean, 'cause that- Yeah ... 'cause that's the other thing, is that in Formula 1, sometimes you'll see another team has done something and they're performing better. And so maybe, w- you know, one team would spot another team does something different, and they go, "Oh, may- well, maybe we should be doing that." And so c- you know, how quickly can you adapt? Or, you know, there's some examples of, of things that aren't to do with the cars. There's... One of my favorite stories is the, is the jack. So when a car comes into the pit box, there's a, there's a jack person who lifts the car up, then they can take all the tires off, and then they drop the car back down, and the car drives off. And one team realized that, um, gravity was actually slowing the car down to get the car back down onto the, onto the track. And so they, they basically put some, I don't know, pneumatic rams or something to push, to push the car off the jack faster than gravity would, because you could find, you know, a hundredth or a thousandth of a second. So I just think that, like, try to find performance in every single element, every single area of your organization. That transfers so much to, to the real world, so to speak. SAM JARRELL The numbers are incredible. I'm sure that's, like, a one data point that they've gotten out of, like, what? I think he said more than seven billion data points from one car. Also, as someone who works for HPE, I couldn't help but appreciate the sort of networking discussion around sending all of that data back and- Yeah ... how important it is to having, like, bandwidth do that in real time. The network is another seriously critical component of the car. And I think he brought up availability and reliability as being so, so important. And when I hear that, it goes back to some points I think we've heard Rami make in the past around how it's, like, not enough for your network to just sort of be there these days or even on or even fast. People's expectations start at it being fast and ready and reliable. So this, you don't notice it when it's working great, but you do notice the second that it's down and not working well. MICHAEL BIRD I mean, could, could you imagine Being at a race, a race weekend, and your, the network going down or- ... your, your IT infrastructure going down. I mean, y- could you just imagine the impact that would have on the race team's performance? I- IT, uh, technology isn't just like a nice-to-have these days. I think we've talked about this on previous episodes. It's like it's critical to your business. Mm-hmm. It's so fascinating how you can kinda keep that network going. And yeah, sending all this data back to the factory, um, making sense of that data, and then using that as this sort of continuous engine to make the car go faster. SAM JARRELL Listen, if it rains on my trek to work, my 30-minute drive- then I'm already having a bad day, and I'm having to figure out how I'm going to get, you know, 10 minutes'- Yeah ... worth of gains back. But I can't imagine also surviving crazy sandstorms, insane environments, while also maintaining such high levels of performance. It really is otherworldly to me. MICHAEL BIRD Yeah. I- it's really impressive, which leads me onto my next question for you, Sam. If you could have the chance to drive an F1 car, uh, w- would you? And if so, is there a particular track or country which you'd want to have a go at racing a car around? SAM JARRELL Well, I would probably wanna try to do the Las Vegas track, 'cause we always have Discover there, and so I feel like it would be fun to go around the same area that we've had our event. What about you? MICHAEL BIRD Uh, yes, I would love to have a go. I would almost certainly crash the car on the first corner. Um, and I would drive Silverstone, which is the track of the British Grand Prix, because it is an absolute classic, and, uh, I just think it'd be amazing to drive around. But yeah, I would almost certainly bin it on the first corner. SAM JARRELL Oh, gosh. Well, we'll stick to our day jobs then. MICHAEL BIRD Yeah, yeah. Oh, yeah. Um, interestingly, this was also the last thing I wanted to ask Michael, as he's been in the industry for a while, so I thought he might have a bit of an insight into where could be a fun or interesting track to try out. MICHAEL TAYLOR This is my 25th season in Formula 1. It's one question that I've never been asked. So, I'm traditionally asked which are my favorite circuits. and generally they're the ones that begin with S. MICHAEL BIRD S MICHAEL TAYLOR And I think it would probably be the same answer to you. I'm not sure whether I would be able to choose between Silverstone or Suzuka. Two amazing- Great tracks drivers tracks. And when you speak to all of the drivers, I know you, you mentioned you interviewed George before. I'm sure George speaks very highly of both locations. They are the ultimate test. High speed, low speed. even medium speed. Linked corners. kinda linked braking zones. So yeah, probably if I had any talent at all, probably one of those two SAM JARRELL Okay that brings us to the end of Technology Now for this week. Thank you to our guest, Michael Taylor And of course, to our listeners. Thank you so much for joining us. MICHAEL BIRD If you’ve enjoyed this episode, please do let us know – rate and review us wherever you listen to episodes and if you want to get in contact with us, send us an email to technology now AT hpe.com Sam, subject line? SAM JARRELL Silverstone or Suzuka MICHAEL BIRD Yeah. Let us know what your favorite race circuit is, and then tell us if you like the show or not. Anyway, don’t forget to subscribe so you can listen first every week. Technology Now is hosted by Sam Jarrell and myself, Michael Bird This episode was produced by Harry Lampert and Eva Higginbotham with production support from Alysha Kempson-Taylor, Nik Damarell, Beckie Bird, Mari Lee, Alissa Mitry, and Jenessa Ayache. Our video editor was Leon Radchinski-Gorman and our theme music was composed by Greg Hooper. SAM JARRELL Our camera team were Branson Tarr, Karl Norberg and Gilbert Rios, Our social editorial team is Rebecca Wissinger, Judy-Anne Goldman and Jacqueline Green and our social media designers are Alejandra Garcia, and Ambar Maldonado. MICHAEL BIRD Technology Now is a Fresh Air Production for Hewlett Packard Enterprise. (and) we’ll see you next week. Cheers! SAM JARRELL Bye y’all