What is Slurm?
Slurm (originally Simple Linux Utility for Resource Management) is an open-source job scheduler used to manage shared computational resources on HPC clusters.
Slurm allows many users to share the same infrastructure fairly and efficiently by:
- Queuing jobs
- Allocating CPU, memory, GPU, and node resources
- Enforcing time and partition limits
- Tracking job history and resource usage
SCU clusters use Slurm to manage compute access across multiple hardware types and lab-specific partitions.
Why Slurm Matters
Instead of running heavy computations directly on login nodes, users submit jobs through Slurm. Slurm then places those jobs onto appropriate compute nodes based on available resources, partition access, requested runtime, memory, CPUs, and GPUs.
Check Current Partition Availability
Partition availability changes over time as nodes are allocated, drained, or taken offline for maintenance. To view the most current state of the cluster, run:
sinfoTo see node-level details including CPU count, memory, runtime limit, and partition, use:
sinfo -N -o "%25N %5c %10m %15l %25R"Notes
- Partitions and node counts may change as hardware is added, drained, or repurposed.
- Some nodes may appear in multiple partitions.
- Preemptible partitions are intended for opportunistic workloads.
- Always use
sinfoon the login node for the most current cluster state.