What is Slurm
Slurm (Simple Linux Utility for Resource Management) is the primary open-source workload manager for HPC and AI clusters. In BCM environments, Slurm is the job scheduler.
Slurm architecture:
| Component | Role |
|---|---|
slurmctld | Central controller daemon — runs on the head node |
slurmd | Compute node daemon — runs on every worker node |
slurmdbd | Database daemon for accounting |
srun / sbatch / salloc | User-facing commands for job submission |
Configuration files:
/etc/slurm/slurm.conf— Main config/etc/slurm/gres.conf— GPU (Generic Resource) config/etc/slurm/cgroup.conf— cgroup settings