Job scheduler
Software that queues jobs on a shared cluster and decides which run where and when.
Also: workload manager, resource manager
A job scheduler takes requests from many users, such as "64 nodes for three days", queues them, and assigns hardware according to priority, fairness and what is free. It starts jobs, watches them and cleans up afterwards. Slurm dominates in supercomputing and is widely used for AI training; Kubernetes, with add-ons for batch work, is common for inference and in cloud-native setups. Good scheduling raises utilisation, and topology-aware schedulers place a job's GPUs close together on the network.