When integrating Ansys HPC Platform Services with a cluster, you can take advantage of the Autoscaling Service.

Like an evaluator, an autoscaler is a form of worker. It requests work from the Job Management Service and uses a scaling strategy to launch workers (evaluators) to meet the demands of different workloads. It then sends jobs to traditional job schedulers (Slurm, LSF, PBS, and UGE), the cloud-native orchestrator Kubernetes, or the local backend. For more information on how this works, see Autoscaling.

When an autoscaler is started, it registers with the Resource Management Service. It will also re-register with the Resource Management Service periodically to look for configuration updates. If any updates are detected, they are applied to the autoscaler configuration file.

Deploying an autoscaling cluster involves configuring and deploying the Autoscaling Service and setting up evaluators for the autoscaler to spawn.

As described in Authentication and Authorization, the Autoscaling Service submits jobs to a cluster as the service user. If you want jobs to run as the submitting users, you must deploy the Process Launcher component.

For a more in-depth look at how the Autoscaling Service works, see Autoscaling Strategies Explained.