Slurm
The open-source job scheduler that runs most of the world's supercomputers and many AI training clusters.
Also: Slurm Workload Manager
Slurm began at Lawrence Livermore National Laboratory in the early 2000s as the Simple Linux Utility for Resource Management. It is open source and is used on a large share of TOP500 systems. Researchers submit batch jobs asking for nodes and GPUs, and Slurm queues and places them. Many AI labs prefer it for training because it is simple and predictable. Neoclouds often offer managed Slurm clusters, and some run Slurm on top of Kubernetes to combine both.