Table of Contents:

Related Articles:

What is Slurm?

Slurm (Simple Linux Utility for Resource Management) is a free and open-source workload manager and job scheduler widely used in high-performance computing (HPC) environments, including many of the world’s top supercomputers. On chip, Slurm manages how computational jobs are submitted, scheduled, and run across the cluster. It performs three primary functions:

• Allocating resources: Grants users access to compute nodes (CPUs, GPUs, memory) for a specified period so they can perform work.

• Job execution and monitoring: Provides a framework for starting, running, and tracking jobs, including parallel and batch workloads.

• Queue management: Arbitrates resource contention by managing a queue of pending jobs and deciding which jobs run when and where.

Slurm is highly scalable, fault-tolerant, and configurable, making it suitable for clusters of all sizes. For more background, see Slurm on Wikipedia or the official Slurm documentation.

How to use Slurm: Basic Commands

You can access resources by interacting with our workload manager, Slurm. Here is more information on Basic Slurm Commands: 

Chip Slurm Usage Rules

Our cluster's compute resources are divided into two categories: chip-cpu and chip-gpu. These are then each subdivided into partitions, in which only certain users can access certain partitions of hardware. Each partition and CPU/GPU requires specific options when using srun or sbatch.

Slurm Accounts vs. Users

It's important to understand the distinction between Slurm accounts and users. An account is a group that a user (you) belongs to.

An overview of which types of accounts can run on what can be found here:

chip Partitions and Usage