Open alahiff opened 2 years ago
For multi-node MPI jobs we possibly should have a single infrastructure containing multiple nodes. This would enable us to properly check quotas etc to properly decide where to deploy it.
For multi-node MPI jobs we possibly should have a single infrastructure containing multiple nodes. This would enable us to properly check quotas etc to properly decide where to deploy it.