Altair and NVIDIA at Argonne: Using Workload Management Software to Maximize HPC Resources

Print Friendly, PDF & Email

Here’s a fascinating look at how a national lab works in partnership with advanced computing vendors to further the lab’s supercomputing mission. From Argonne National Laboratory’s leadership computing facility (ALCF) we have Bill Allcock, manager of the Advanced Integration Group, and Bill Nitzberg, CTO of PBS Works at Altair Engineering.

The two talk about ALCF’s decision to retire its in-house Cobalt workload manager in favor of Altair’s PBS Professional and how the software will be used to optimize the facility’s HPC systems – including the Polaris testbed supercomputer and the upcoming Aurora exascale system – to accelerate science and engineering research.

The goal in selecting PBS Professional is to optimize the complexities of managing diverse workloads, including machine learning and data-intensive tasks on the Polaris system, which utilizes 2,240 NVIDIA A100 Tensor Core GPUs.