Standing reservations for short jobs

Generally, the longest amount of time an individual job may run on SciClone and Chesapeake is either 72 or 180 hours, depending on the subcluster. This allows for calculations that may take several days to complete to run without needing to be subdivided into shorter jobs, but also means that even if you are next in the queue, you may have to wait several days before nodes become available.

To facilitate shorter turnaround for debugging and test jobs (or any other short jobs), we maintain standing reservations that further restrict maximum walltime on a few nodes:

Subcluster Number of nodes (cores) reserved
Longest walltime allowed
When active
Whirlwind 2 (16) 1 hour 8am-6pm M-F
Hurricane 2 (16) 1 hour 8am-6pm M-F
Whirlwind 4 (32) 3 hours 8am-6pm M-F
Vortex 2 (24) 1 hour 8am-6pm M-F
Vortex 6 (72) 8 hours

7:30am-3:30pm and

7:30pm-3:30am every day

Jobs whose resource specification meets the restriction will automatically be able to use the reserved nodes. Please contact the HPC group if you have any problems using these reservations, or if you need a reservation on another subcluster.