Slurm is a scheduler. It is the software that you install on a bunch of GPU servers to make them into a cluster.
Instead of having users login to individual machines in this cluster, users now get to use slurm commands to launch batch jobs (“sbatch”) or allocate a node for interactive work (“salloc”, “srun”).
Slurm decides what job runs on what GPUs at what time. You can define priorities/partitions based on resources, track what users ran what jobs (“sacct”), and store error/output logs simply.
If you’re visualizing everything as a stack, slurm goes on top of your operating system, but below your PyTorch code.
Nov 6 • 16 tweets • 7 min read
ClusterMAX 2.0 is here!
In this update we rate 84 neoclouds, up from 26
Our market view tracks 209 providers, up from 169 in March and 124 last October
We spoke to over 140 end users as part of the research
And we wrote over 43,000 words about our experience
This is more than Animal Farm, but less than The Great Gatsby
