Scheduling

A means to take requests for work and order them in a way by priority (fair share or other algorithms) and availability. Queuing systems are interested in attributes such as satisfiability, reservations, and backfilling.

Example

In Kubernetes, a tool like Kueue [ref] might create simple job requests for users and assign them to be run based on priority settings. In an HPC resource manager like SLURM [ref], it’s more likely that groups of nodes (called partitions) are allocated to user jobs based on an algorithm like fair share.




Suggested Citation:
Converged Computing Authors. "Scheduling." Converged Computing Community Space 29 Nov 2024, https://converged-computing.org/docs/queueing/ (accessed 29 Nov 24). [doi]