Slurm Accounting in PCS gives your scheduler a memory

In this Tech Shorts episode, Matt Vaughn returns to dive into the power of Slurm accounting in AWS PCS

Why it matters, how it gives clusters “memory” of past workloads, and how that transforms scheduling, efficiency, and user fairness.

From tracking GPU usage and setting quotas, to enabling policies that balance entire departments’ needs, Slurm accounting unlocks smarter resource allocation and deeper insights.

The best part? PCS makes it as simple as a checkbox—no messy databases to manage, no fragile setups. Whether you’re curious about HPC cost control, performance optimization, or building sophisticated policies for your teams, this episode shows how accounting turns Slurm into something far more powerful than a scheduler.

If you have ideas for technical topics you’d like to see us cover in a future show, let us know by emailing us at ask-hpc@amazon.com.

Chapters

00:00 - Start 02:00 - Retrospective accounting 02:52 - Forward looking use cases 03:58 - Consumption control 06:52 - How does Slurm Accounting work in PCS? 09:57 - What metrics does it track? 11:59 - How do you set quotas? 15:18 - How to play with Slurm Accounting