The HPC that does not get a vacation

The Caterham F1 Team raced in the 2013 Singapore Grand Prix last month. Credit: Dell

The Caterham F1 Team raced in the 2013 Singapore Grand Prix last month. Dell’s High Performance Computing (HPC) Cluster can do 10 billion calculations in 17 hours, says William Morrison, Caterham IT infrastructure manager. Credit: Dell

The obsession with saving time is not just confined to the race track in Formula 1TM. Getting the most performance from every part is vital to making a car quicker, and parts are constantly improved and updated throughout the season in the quest to shave tenths of a second off the lap time. Understanding aerodynamics is the key to unlocking speed … and that’s where Caterham F1 Team’s  Dell High Performance Computing Cluster (HPC), powered by Intel, comes in.

The Dell HPC that Caterham has in their Leafield Technology Centre in Oxfordshire, England, is a sight to behold — but it seems even more incredible when you realize what it actually does.

Complex calculations

Formula 1 teams have long relied on wind tunnels for testing parts to find out if they are going to add performance to the car or not, yet the advance of technology has allowed F1 teams to make use of computational fluid dynamics (CFD) in tandem with wind tunnel work. CFD starts with a process called meshing, which is akin to draping a fishing net over the car. The wind tunnel inside the computer will then give Caterham’s CFD engineers detailed information on what is happening with the temperature, pressure, turbulence and velocity inside each tiny triangle in the mesh that covers the surface of the virtual car.

The process of trying out virtual parts before they are put through to the next stage and physically made is vital in saving that precious commodity: time. “By trying out a number of different ideas and testing whether they will work or not, it means we are able to do development work without actually having to make parts,” explains William Morrison, Caterham F1’s IT infrastructure manager.

“For an average 17-hour job the HPC will do approximately 10 billion calculations,” Morrison says. That is the primary function of this supercomputer. To put that in context, the average PC you might have at home to surf the Internet would take between four and five months to get through that amount of math.

Once the HPC has performed its calculations, it will streamline these to about 800 million pieces of individual data, which will then be presented to the CFD analysts in the form of approximately eight videos, a handful of graphs and a couple of hundred pictures.

Time-saving enhancements

Behind the scenes, however, there is always work being carried out in connection with the HPC in order to make it quicker and more reliable; all linking back to that vital factor of time.

“We recently got the time of a typical job down to about 12 hours,” Morrison says. “This was done through enhancements to the model set-up, and optimizing the way the model solves. We have a little group of about three people dedicated to improving this all the time.”

Shaving this five hours off the time of a solitary job was the result of several months’ work, stripping the model back and rebuilding it with different numeric and solver settings, which all needed testing and correlating.

In case of emergency

Like any piece of customized computing equipment of this nature — the HPC requires careful upkeep. Though it does not require physical upkeep, the HPC is monitored twice a day and gives the team the ability to take individual nodes offline without affecting other jobs.

So given the importance to the team of the work constantly going through the HPC, what happens in the event of an unwanted event like a power outage?

“We have battery conditioning units so that there’s always conditioned power coming into
the HPC, and if there’s a disruption to the external power supply we’ve got a generator outside which kicks in automatically,” says Morrison.

With the quest for saving time the constant motivation, the Dell HPC is the nerve center of the entire Caterham F1 Team operation. Without it their car simply could not be designed or developed, which is why the machine runs 24 hours a day, 365 days a year — always with a queue of between 10 and 20 jobs waiting to be solved by its mammoth powers of calculation.

“It’s not allowed any time off,” laughs Morrison.

Want more? Stay up to date on the latest news and events about the Dell/Caterham F1 Team partnership. Follow us on twitter @dellinsidetrack, and read the other parts of the series:

Also in Dell-F1 Series 2

Dell and the Caterham F1 team provide an inside look into the powerful enterprise and trackside infrastructure that lays the foundation for racing success.

  1. 1A ‘half rack’ is critical to racing success
  2. 2The HPC that does not get a vacation
  3. 3Weight of an F1 team sits on one man’s shoulders
  4. 4An unsung hero of Caterham F1 team

View the entire series.

Hans Seeberg
Hans has been a magazine journalist for 17 years, with particular experience in motoring, men’s lifestyle and Formula 1.
Hans Seeberg
Tags: Data Center,Technology,Virtualization