Institutional allocations and job scheduling/preemption

Created by Robert Harrison, Modified on Fri, 14 Aug at 11:10 AM by Izabel Cavassim

All queues across the Alpha and Beta clusters use a fair-share scheduling policy. For more information check our Fair Share and Performance Model.


How slurms fair share works


Fair share is a scheduling algorithm that adjusts job priority based on historical resource use. Users or groups who use fewer resources get higher priority, while those who use more get lower priority to keep the system fair for everyone.


Shares: An assigned allocation or weight given to a user or project group based on contribution or policy.

Usage: The amount of compute time (CPU, memory, GPUs) consumed over time.

Half-Life Decay: Past usage fades over time using a set half-life (currently set at 14 days), meaning recent jobs matter more than old jobs.

Fair-Tree Algorithm: Modern Slurm versions use a tree structure algorithm to compare usage and shares hierarchically across accounts and sibling groups.


Checking your status


Run sshare in your terminal to view current account shares, raw usage, and calculated fair-share factors.

  • Values closer to 1.0 mean low resource use and high priority boost; values closer to 0.0 mean heavy use and lower queue priority.



Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select at least one of the reasons

Feedback sent

We appreciate your effort and will try to fix the article