// NCCS Quality of Service Details

Slurm's Quality of Service (QoS) feature controls resource limits for every job in the Discover job queue. Available QoSs in the table below apply only to jobs submitted to the Slurm default partition. (It is important for maximum adaptability of your job scripts that you not specify any partition if you wish to use the default.)

Quality of Service Time Limit Max CPUs/user Max jobs/user
allnccs 12hrs 9000 25
debug 1 hr 1600 1
long 24 hrs 560 25
serial 12 hrs 4116 1

QoS "allnccs" is the default QoS when no other QoS is specified. For maximum adaptability to future changes, users should not specify --qos=allnccs. Use sbatch command-line options or in-line directives to access debug, long, and serial QoSs, by specifying, for example:

#SBATCH --qos=long

Note: unless you specify a time limit, the default is 1 hour. NCCS strongly recommends that you specify as accurate a time limit as possible. See a couple of Slurm approaches to specifying time limits, and other recommended methods for enhancing your job turnaround times, in Methods for Job Turnaround Optimization Using Slurm.

Note: On Discover we do not "time-share" compute nodes, i.e., a compute node can only be used by one user and one batch job at a time. Therefore, please pay attention to efficient usage of the computational resources.