Graeme Hackland is no stranger to the need for speed when it comes to moving data around – he’s the CIO of Williams Racing, a leading Formula 1 team and engineering company. Hackland spoke at CloudFest to share some of the amazing data science that goes into achieving Formula 1 racing success.
Williams started in 1977, said Hackland: “they were a very small team, taking on the bigger teams,” and today’s mission is to repeat that. Furthermore, Williams uses the tech it develops to forward other industries: did you know that the current F1 car is hybrid-electric? That same Williams battery tech powers those foldable bikes you might see on the streets of your city. Even the curve of the F1 car’s wing has been applied to fridges: an aerofoil keeps cold air in!
Making the right decision more often
Data management is all about making the right decision more often, said Hackland:
William’s path to success began when it started using data.
Back in the day, it took 20 minutes to download one lap’s worth of data… today, a car will generate 800GB per race. The question became, how can you get access to this data in real time?
(By the way, F1’s situation is writ large across the cloud industry as a whole: Cisco forecasts global cloud datacenter traffic to hit 19.05 Zettabytes by 2021 – which means it will have nearly doubled since last year. Its value in 2021, according to Gartner, will be $83.5 billion by 2021. A zettabyte, by the way, is 1,000,000,000,000,000MB.)
Graeme Hackland: the car has over 1000 channels of data, but there were none on the driver
“I think we’re in a really good phase in Formula 1,” said Graeme Hackland, and that’s partly down to the driver – Hackland isn’t looking forward to something like driverless F1. However, the car has over 1000 channels of data, but there were none on the driver. That’s changing, though, to make medical teams’ jobs easier; and the benefit goes much further. Human biometric data, said Hackland, is another tool to help the team make better decisions before and during each race.
To get that car on the track, the factory itself has to generate and make sense data using hardware such as wind tunnels and supercomputers. “Aerodynamics is where the most data is generated,” said Hackland, but a lot of it has to get dumped: “If we could have infinite storage, we would!” Manufacturing data is also very valuable, he said, for both speed and iteration.
All that data has a sell-by date, said Hackland: only valuable for a short period of time. So where is the valuable data, and for how long before it’s no longer relevant? His team puts a lot of time and effort into the data, its structure, and its integrity. Using the cloud during the race was very attractive, said Hackland: for the last five years Williams has been running computation in the cloud. Engineers used to run data on their laptops for hours at a time in the evening – now those computations can get done while the car is still on the track. Obviously, he added, such processing is not a cost-saving exercise: “we justified it on what we would be able to do.” That said, Hackland noted that the cost avoidance on lugging servers all over the world actually paid for their virtualization rollout.
“Your competitors are in the garage next door to you,” said Graeme Hackland. “It’s very public: when you don’t get it right, the public knows!”