Queues on PLX - update

Dear Users,

the modification of the queues configuration on PLX was successfully implemented with no change in the schedule of the queued jobs. The modification aimed at easing the workload of the parallel queue, making it as responsive as the longpar queue. The two classes are now equivalent, and whatever the queue you ask for, either longpar or parallel, all jobs will result to be served by a unified, longpar list. Please, note that:

• the parallel class is still defined and you can keep requesting it in your submission scripts (by the #PBS -q parallel directive). Hence, no change in your parallel job scripts is required.
• the restriction on the six hours maximum walltime existing for the parallel queue is no more applied, and for both the parallel and the longpar queues a walltime of 24 hours can be requested. However, we strongly encourage you to limit the requested walltime to the strictly necessary amount (for both parallel and longpar queues) and, whenever possibile, to partition your jobs in chains of shorter runs: the shorter the requested walltime is, the more optimized the performance of the scheduling will be, as your jobs will benefit from the backfilling action which will reduce the overall queueing times.

We wish you a happy holiday season,

best regards

HPC User Support @ CINECA