German engines manufacturer MTU Aero Engines says it’s getting very high utilisation and prioritised simulations from its Platform LSF high performance computing (HPC) installation.
The company, which makes commercial and military engines, is currently upgrading to the latest version (Platform LSF 7), primarily for its R&D, and says it expects that investment to secure growth until 2012.
That’s on a large Linux cluster in Germany and a smaller one at a subsidiary in North America, with around 150 users submitting an average of 3,000 jobs a day. LSF intelligently schedules parallel and serial workloads, such as Abaqus, and MSC.Nastran, to solve engineering problems.
Most of the CPUs are dual or quad core, though some older single core CPUs remain. The company says it has maintained a mix of hosts because some of its applications cannot take advantage of newer multi-core CPUs, while others request big memory.
According to Axel Philipp, MTU’s IT administrator, utilisation of its 1,400 CPUs in the main cluster averages 80%, with some months taking that to more than 90%.
“With Platform LSF we can achieve high utilisation while still getting free slots for those who really need them,” he says. “That is what I need Platform LSF for – to do fair share scheduling to prioritise jobs and pre-empt those that are less important.”
He also makes the point that, while most jobs submitted to the cluster are long running, the cluster also allows him to accommodate quick jobs. “We reserve some licenses, during the day time, for short queues, so that the users can get their 20-minute jobs through quickly,” says Philipp.
And he adds that using the HPC technology has enabled MTU to defer hardware acquisitions, estimating that “at least 10% more hardware capacity [would be needed] without the scheduling possibilities of Platform LSF.”