> Yesterday I played with SR. It also works with single jobs, > not with multi-cpu/multi-gpu jobs and don't protect from > occupation of an one CPU per node for gpu jobs.
We have SRs on our 128p shared mem system which 'protect' only some (floating) cpus. We are using moab and have SR's like: # set aside some resource _not_ for really really long jobs SRCFG[hour24] TASKCOUNT=1 MAXTIME=24:00:00 SRCFG[hour24] RESOURCES=PROC:8 SRCFG[hour24] PERIOD=WEEK DEPTH=2 SRCFG[hour24] FLAGS=ACLOVERLAP,IGNSTATE I think this idea should be reproducible on a multi-node cluster. -- Gareth _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
