Slurm is the management and job scheduling system for the Zorbas cluster. Slurm’s primary role is to allocate access to resources (computing nodes) to users for some duration of time and to provide the framework for starting, executing and monitoring the jobs on the set of the allocated nodes. A slurm job can be a single command (in any scripting language like bash, awk, python, perl, R etc) or a complicated set (pipeline) of commands.
The entities managed by slurm are nodes, partitions, jobs and jobs steps:
- Nodes: computing units with specific number of cores and amount of memory
- Partitions: logical node sets with similar specifications
- Job: an allocation of resources assigned to a user
- Job step: a parallel task within a job ( A job may consists of one or more parallel steps)
For more information, please refer to the official website of slurm workload manager.
For a brief summary of slurm commands and options please download this pdf file.