Overview

SeaWulf uses a fairshare scheduling system to allocate resources. Job priority is determined by factors such as historical resource usage, job size, and requested time. Understanding this system helps maximize throughput and reduce wait times.

Key Concepts

• Fairshare: Users who have used fewer resources recently get higher priority for job scheduling.
• Job Priority: Factors include queue type, number of nodes/cores requested, walltime, and fairshare score.
• Preemption: Some queues may allow higher-priority jobs to start before lower-priority ones.

Tips for Maximizing Throughput

• Submit smaller jobs when possible; many small jobs often start faster than one very large job.
• Use appropriate queues; match your job to the queue’s walltime and node limits.
• Stagger job submissions to avoid all jobs entering the queue at the same time.
• Monitor your job’s fairshare score using squeue -u <username> or other SLURM commands.
• Consider using job arrays for repeated tasks to reduce queue overhead.