Myrinet v/s infiniband nodes

To control the placement of your job on the Myrinet (older) or Infiniband (newer) part of the cluster, you can specify the interconnect=mx or interconnect=fdr option in your resource limits specification.

$ qsub -I -l select=1:ncpus=2:mem=2gb:interconnect=mx,walltime=1:00:00

Specify the phase

You may specify a particular phase on which your jobs should run, using the phase= option in your resource limits specification.

$ qsub -I -l select=1:ncpus=2:mem=2gb:phase=8c,walltime=1:00:00

Specify the host

Similarly, you can have your jobs land on a specific node:

$ qsub -I -l select=1:ncpus=2:mem=2gb:host=node1744+1:ncpus=2:mem=2gb:node1745,walltime=1:00:00

Or, in the case of multiple hardware chunks, you can specify a node for each chunk:

$ qsub -I -l select=1:ncpus=2:mem=2gb:host=node1105+1:ncpus=2:mem=2gb:host=node1106,walltime=1:00:00
$ cat $PBS_NODEFILE
node1105.palmetto.clemson.edu
node1106.palmetto.clemson.edu

Controlling the placement of resource chunks

When requesting multiple chunks of hardware, you may force each chunk to be on a different node or on the same node, using place=scatter or place=pack:

-l select=2:ncpus=2:mem=15gb,walltime=00:20:00,place=scatter    # force each chunk to be on a different node
-l select=2:ncpus=2:mem=15gb,walltime=00:20:00,place=pack       # force each chunk to be on the same node