Using partitions
In general, it is not needed to choose a partition for your job, as this is done for you by the job modifier script depending on the project and resources you are asking for.
Every project has a default partition and (eventuelly) additional partitions, it is allowed to to submit to.
partition | #nodes | #cores per node | #mem per node | billing | remarks |
c18m | 1240 | 48 | 192 | cpu=1 mem=0.2 | default partition for default project |
c18g | 54 | 48 | 192 | cpu=1 mem=0.25 gpu=24 | 2 V100 gpus, request of volta gpu needed to submit to this partition |
c16m | 608 | 24 | 128 | cpu=1 mem=0.1875 | project needed to be able to submit to |
c16s | 8 | 144 | 1024 | cpu=1 mem=0.140625 | project needed to be able to submit to |
c16g | 9 | 24 | 128 | cpu=1 mem=0.1875 gpu=12 | 2 P100 gpus, request of pascal gpu needed to submit to this partition |
Explanation of the billing values of the table:
cpu is clear, one cpu equals the requested cores
mem means cores per gigabyte, for c18m e.g. 1 GB requested memory equals to 0.25 cores, or 4 GB equal to 1 core
gpu: there are only gpu jobs allowed on the gpu nodes, each node has 2 gpus, thus 1 gpu equals to a half node