Description
The national systems use a job scheduler called “Slurm”. In this session you learn how Slurm works and how it allocates jobs, helping you to: minimize wait time by framing reasonable requests; ask for only the resources you need to improve efficiency; increase throughput; run more jobs simultaneously; and troubleshoot and address crashes.
This workshop is designed for either new high performance computing (HPC) users familiar with Linux and Shell Scripting, but who have not had experience using Slurm, or, for experienced users transitioning to Slurm or seeking to improve efficiency with the scheduler.
Prerequisites
Introduction to Linux and Introduction to Shell Scripting, or equivalent experience. Participants should have an active CCDB account.
Details
Learning Objectives
- Describe how Slurm schedules jobs on HPC systems
- Submit jobs using Slurm commands and scripts
- Request resources such as CPU, memory, and time for jobs
- Improve job requests to reduce wait times and increase efficiency
- Run multiple jobs at the same time using Slurm
- Check and manage running and queued jobs
- Identify and fix common job errors
Who Should Enrol
This workshop is intended for HPC users who are comfortable with Linux and basic shell scripting but are new to Slurm, as well as experienced users looking to transition to Slurm or improve their efficiency with the scheduler.