Factors that Determine Job Priority:

  • Age - the amount of time the job has been waiting in the queue
  • Job Size - number of nodes requested by the job
  • Partition - priority for a given partition
  • Fairshare - priority contribution based on compute resources used by members in a research group within the last 30 days.
    • The more jobs that have run on the cluster the lower the priority.
    • The fewer number of jobs that have run the higher the priority.
  • Quality of Service (QOS) - supporters of CCR are given a priority boost for their group's jobs.  Find out how to become a CCR supporter


Click on the links for more details about SLURM Multifactor Priority Calculations and Fairshare



Show Fairshare:

sshare <flag>

--accounts=group_name

--all --accounts=group_name  (Shows fairshare for members of the group)

See SLURM Account Information for group_name (account)



Job Priority Formula:

Job_priority =
(PriorityWeightAge) * (age_factor) +
(PriorityWeightFairshare) * (fair-share_factor) +
(PriorityWeightJobSize) * (job_size_factor) +
(PriorityWeightPartition) * (partition_factor) +
(PriorityWeightQOS) * (QOS_factor)



Show Job Priority:

sprio <flag>
-j jobid
-u username



Show Job Priority sorted from Highest to Lowest:

sranks




Weights assigned to Priority Factors for UB-HPC cluster:


PriorityWeightAge       = 50000
PriorityWeightAssoc     = 0
PriorityWeightFairShare = 80000
PriorityWeightJobSize   = 200000
PriorityWeightPartition = 1000000
PriorityWeightQOS       = 50000
PriorityWeightTRES      = CPU=0,Mem=.01,GRES/gpu=30000