Costs & Scaling

What do I pay for? You pay for the head node, compute nodes (only while running), and storage. ParallelCluster itself is free — you only pay for the underlying AWS resources.

What runs when no jobs are submitted? Only the head node and shared storage. Compute nodes scale to zero when idle (if MinCount is 0).

How do I minimize costs?

Set MinCount: 0 on all compute queues so nodes scale to zero
Delete your cluster when you’re done for the day/week
Use the smallest head node that meets your needs
Back up results to S3 and delete the cluster between sessions

How does auto-scaling work? When you submit a job, Slurm signals ParallelCluster to launch the required instances. After jobs complete, instances stay idle for the cooldown period (default 10 minutes), then shut down automatically.

Can I change the idle timeout? Yes. Set ScaledownIdletime in your cluster configuration under Scheduling > SlurmSettings.

How much does a typical cluster cost? It depends on instance types and usage. See the Cost Estimates Reference for examples.