Because the Keck Cluster is a shared resource, a job scheduler is used to provide fair access to all cluster users.
-
The scheduler used on the Keck Cluster is SLURM.
-
Users on the cluster submit requests for resources to the cluster scheduler
-
If the scheduler finds a block of resources large enough to satisfy the user’s request, it allocates those resources to the user immediately
-
If there aren’t enough resources free when the request is submitted, the user’s submission is placed in the queue specified in the user’s request
-
On the Keck Cluster,
SLURMis configured to use a history-aware weighted-fair-queuing algorithm to decide which queued requests to run first -
SLURMcan be used to request interactive sessions or to submit batch jobs
Documentation Links
SLURM Scheduler Quick Start Guide
SLURM Scheduler Video Tutorials
-
There are many additional resources on the web for learning
SLURMconcepts and the syntax ofSLURMjob-submission scripts:
Here is one to start with (in English, from CÉCI, a national consortium of HPC centers in Belgium):
Slurm Quick Start Tutorial
-
Official documentation for the version of
SLURMcurrently in use on the Keck Cluster:
-
Documentation specifically for the
SLURMsbatchutility and job-script syntax:
Contact Us
Contact researchcomputing@chapman.edu with questions or requests.