Formula One Grand Prix motor racing is a prime example of The Engineer’s Curse, which states that the closer they come to perfection in their job, the easier it looks. Many don’t appreciate the magnitude of the scientific and technical advances made in F1 because it’s not enough of a visual spectacle, unless someone crashes or there’s some drama in the pit stop. In fact, those two-second pit stops are like the entire industry in microcosm. Behind that one car, there are up to 700 scientists, data analysts, engineers and all grades of technician frantically working behind the scenes to surmount massive logistical challenges to impossible deadlines.
All the real drama is in the development labs. Red Bull Racing, being a spinoff from a drinks company, had a lot of catching up to do in an industry where all the competitors like McLaren, Ferrari and Renault have centuries of engineering heritage between them. The fact that Red Bull Racing has finished in the first three in half of its races means that its intellectual capacity has roared around an impressive learning curve. Arguably, it is high-performance computing vehicles that drive it, fuelled on high-grade data.
The computing engines that power the England-based Red Bull Racing team to victory over their massive logistical challenges are managed using Software Defined Computing and Storage by IBM. It provides a stable of platforms that includes the Analytics, Application Center, License Scheduler, Process Manager, Symphony and the General Parallel File System (GPFS) models.
Like any good system of engineering, these are designed to make the task of re-inventing the car run so smoothly you could be fooled into thinking it is easy. However, the statistics for each Grand Prix competition it enters illustrate the scale of the task.
This year's model for the Red Bull Racing team, the RB12, has 100 sensors embedded in its body. Each sensor gathers data on a particular aspect of the machine’s performance, such as tyre temperature variance, the oil levels and pressure, the wind resistance experienced on various parts of the engine. All these elements affect the performance of the car. If they can be tweaked to make even a millisecond’s worth of improvement to lap times, the aggregation of all these changes could add up to the difference between first and second place.
The intelligence from these sensors is constantly reported back which, in the interests of creating a level playing field, the race regulation body FIA (Fédération Internationale de l’Automobile) restricts to 10 gigabits per second for all participating teams.
The data accumulated from all these reports flows into the IBM General Parallel File System. For competitive reasons IBM and Red Bull won’t specify how much data but Petabytes were mentioned on the media tour of the RBR HQ in Milton Keynes I joined a few weeks ago. Those mountains of IoT information, often in the form of unstructured data, must be analysed and acted upon immediately, so teams of data scientists will be working on key applications running on the IBM managed super computers with all the urgency of a pit stop wheel change team, anxious to diagnose any unusual traits in the machine’s performance.
Once the race is over, the real work begins for most of the 700 staff in the Red Bull Racing team. So begins an eight-day cycle of continuous design interaction, simulation and testing, using the data about the car’s performance in the previous race as a guide to redesigning the new model.
There are 21 race events a year in locations all around the world, so the IT platforms will accumulate huge volumes of data created by all the reporting sensors.
In the course of this season the entire RB12 model car could well be redesigned four times over, according to Red Bull Racing’s CIO, Matt Cadieux. On average there are 1,000 redesigns of aspects of the car (ranging from the spoilers to spark plugs, bodywork to brakes) every week, with one million job movements (IBM-speak for mini projects). In total, there are 30,000 design changes created on the car every season and the number rises every year, while the budget stays constant.
The reason all these changes can be accommodated, and the logistics of modelling new parts and getting them manufactured, tested and shipped to the destination of each race meeting, is that IBM’s Platform Process Manager provides a stable framework around which all these complicated tasks run.
If the IBM Platforms were a Formula One car, the sudden acceleration they have produced in the Red Bull Racing team would have the regulators asking all sorts of questions. There has been a 30 per cent reduction in the ‘lap time’ of the design team each time they have to create a new version of RB12, according to Nathan Sykes, CFD and FEA Tools Group Team Leader at Red Bull Racing. The explanation is that workloads used to be set off manually, forcing Sykes’ team to wait until one processor was finished before moving on to the next. This time and resource consuming process was engineered out by the IBM Platform LSF family which can handle simultaneous simulations and the design of complex, interdependent workflows. “Now we can redesign, model and build a new Formula One car in about half the time that it used to take in the past,” says Sykes.
Phil Muncaster reports on China and beyond