Spread work across Compute nodes through a template? RRS feed

  • Question

  • I am running an HPC cluster. The cluster consists of 3 compute nodes, each with 16 cores and 128GB RAM.
    Normally when a user submits a job to the cluster, it assigns all the cores from Compute Node 1(CN1). I would like to create a template so that when a job is submitted, it spreads the cores out horizontally across all the compute nodes. So if a user submits a job which is going to use 12 cores, it would use 4 cores on each CN instead of all 12 cores from the same CN. Is this possible? If so, how would I set this up as a template?
    Friday, July 22, 2011 4:09 PM