System partitions

System partitions

To determine which partitions (queues) exist on a cluster you can use the sinfo -s command. Below is the command output from Aphroditi.

[root@front01 ~]# sinfo -s
PARTITION AVAIL  TIMELIMIT   NODES(A/I/O/T) NODELIST
milan        up 1-00:00:00        26/5/0/31 ne[05-35]
genoa        up 1-00:00:00        11/6/0/17 ne[37-53]
cpu*         up 1-00:00:00        12/0/3/15 cn[03-17]
a100         up 1-00:00:00          2/3/1/6 sim[01-06]
gpu          up 1-00:00:00          2/1/1/4 gpu[03-06]
rtx          up 1-00:00:00          1/3/0/4 rtx[6003-6006]
b200         up 1-00:00:00          1/1/0/2 b[201-202]

The command shows the partition name, status, maximum time limit, node availability and node list.

  • cpu: default CPU nodes partition
  • gpu: legacy Cyclone GPU nvidia v100 nodes partition (to be deprecated soon)
  • a100: GPU nvidia a100 nodes partition
  • milan: AMD milan nodes partition
  • genoa : AMD genoa nodes partition
  • rtx: GPU RTX 6000 PRO blackwell edition nodes partition
  • b200: GPU B200 nodes partition

Applications submitted via the continuous open-access process can be allocated resources across the infrastructure, including the advanced b200 and rtx partitions. Unless specified in the appliation, access will be granted to the lower tier partitions.