Overview

Efficient use of compute nodes on SeaWulf ensures fair access for all users and maximizes throughput. Proper resource management helps avoid wasted CPU time, memory, and GPU usage.

Best Practices

• Match resources to your job: Request only the number of cores, nodes, memory, and GPUs needed.
• Use job arrays: For multiple similar tasks, use SLURM job arrays instead of submitting many individual jobs.
• Monitor resource usage: Track memory, CPU, and GPU usage to avoid over- or under-allocation. See Monitoring Resource Usage on SeaWulf.
• Avoid idle nodes: Release nodes promptly when your job finishes to free them for other users.
• Use efficient code: Optimize your code for parallelism and memory usage to reduce wasted resources.

Interactive Jobs

For testing and debugging, use interactive sessions with srun --pty bash instead of running long computations on login nodes. This ensures jobs run on the correct resources and do not affect other users.