PGS has used best-practice execution of batch and high-performance computing workloads on Google Kubernetes Engine (GKE) to replace 260 000 core Cray supercomputers in Houston.
PGS Chief Enterprise Architect Louis Bailleul and Google's Maciej Rozacki recently presented at the Google Cloud Next ’22 event in Munich on the PGS journey into cloud-based supercomputing. In his latest Industry Insights post, Andrew Long examines the “move and improve” journey PGS has undertaken from on-premise Cray supercomputers to a GKE supercomputer in the cloud.
At peak capability, PGS is using the GKE supercomputer running in the Google Cloud Platform with a hypothetical 72.02 petaFLOPS capability, equivalent to the world's 7th largest computer.
Citing various examples, Andrew points to both the flexibility and the sustainability merits of this approach; Google Cloud Platform is powered by 97% carbon-free energy. Although adapting to the cloud is cost-efficient, it is a significant engineering undertaking, he explains.
Supercomputing in the Cloud
PGS runs large workloads with over 1 million virtual CPUs on multiple GKE clusters. Examples demonstrate the flexibility and significant sustainability merits of the Google Cloud Platform approach.
