Slurm is the scheduling system that we have implemented on

As our main cluster has grown more heterogeneous over time, the PBS/Torque Maui scheduler has not been able to keep up. More specifically, Maui doesn’t seem to want to do the things we tell it do - like run jobs on the faster cluster nodes by default (instead of whevever it chooses). The Maui codebase is somewhat neglected, and it mainly serves as a free version of the Moab scheduler. Slurm also has more redundancy baked in, so if the head node goes down - another can immediately take over.

There is a quick reference for making the change to Slurm, and we’re updating documentation to reflect Slurm script differences.