Following the release of Appro’s commitment to Renault’s Formula1 team, Red Bull Racing has released details on its own HPC infrastructure. The Red Bull Formula1 race team, a major competitor of Renault, is running a 1,024 core IBM cluster with AMD silicon in order to perform their CFD modeling. Red Bull and their sister team, Scuderia Toro Rosso F1, found that with larger core counts and more users, the black art of batch scheduling came into play.
The problem emerged when the guys would set a big job to run overnight,” [Steve Nevey, business development director of Red Bull Technology] said. “Being able to run bigger jobs faster meant they would have to hang around to kick off another job. We needed to automate this with software to handle the scheduling of CFD jobs.”
Red Bull pulled in the help of Platform Computing and their Load Sharing Facility batch software to assist with the system management. As a result, they’ve been able to significantly improve the utilization of the machine.
We’ve been able to optimise the AMD/IBM cluster implemented last year to run CFD jobs by 20 per cent,” he said.
Nevey also mentioned that with a better handle on machine utilization, the Red Bull team plans on adding more structural and vehicle dynamics modelling to their list of workloads.
Read the full post here.
As an avid Formula1 race fan, I try to keep track of the current state of R&D amongst the teams. So far, there are several teams that have publicly announced the use of large-scale HPC environments specifically for their Formula 1 pursuits. Ferrari bought a large chunk of Rackable compute, Renault runs Appro, Honda runs SGI and Red Bull/Scuderia runs IBM.