Parallelization within a compute node

Torc attempts to maximize parallelization of jobs on a single node based on the job resource requirement definitions. Be aware of the fact that the default number CPUs for a job is one, and so it is critical that you define these values conservatively. Refer to Resource Requirements for more information.

If all jobs have similar resource requirements then you can set the option --max-parallel-jobs in the torc hpc slurm schedule-nodes command and avoid having to define the job requirements. Torc will use that parameter to limit concurrent jobs on each compute node.