Slurm Resource Options Explained: Tasks, CPUs and Nodes in HPC Jobs
The article explains how Slurm handles CPU and memory requests when jobs are submitted to high-performance computing clusters. It defines three core concepts: a node as a single computer in the cluster, a CPU core as a processing unit, and a task as an individual unit of work, often an MPI process. The --ntasks option sets the total number of tasks, which Slurm can spread across several nodes. The --cpus-per-task option sets how many cores each task receives, which suits multithreaded applications such as OpenMP, and --ntasks-per-node caps how many tasks run on each node. The author stresses that requesting the wrong resources can waste capacity or delay jobs in the queue, and notes that memory requests such as --mem and --mem-per-cpu are covered separately.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in