In anticipation of more users being added to the Hex cluster we have modified the scheduling rules. Previously there was an open policy:
USERCFG[DEFAULT] MAXJOB=180
USERCFG[DEFAULT] MAXPROC=180
USERCFG[DEFAULT] MAXNODE=180
This effectively allowed users to consume 66% of the cluster at a time. We have now implemented the following hard and soft limits
USERCFG[DEFAULT] MAXPROC=64,180
We've removed the MAXJOB and the MAXNODE directives as we're only interested in restricting access to cores. The implication of MAXPROC=60,180 is that in a heavily used state users will be limited to 60 cores each maximum, i.e. 25% of the cluster, but if the cluster is not heavily utilized the hard limit will be used granting 75% of the cluster to a single user. (See Adaptive Computing's throttling policies for more info)