How Expanse Works

Expanse acts as an intelligence layer on top of your HPC or GPU cluster.

Its purpose is to help clusters run more workloads successfully without requiring additional hardware.

​

Predicting resource fit

When workloads are submitted to a cluster, selecting the right resources is not always straightforward. Jobs may request more resources than necessary or be scheduled inefficiently.

Expanse predicts resource fit at submission time, helping identify how well a job matches the available resources in the cluster.

Better matching between jobs and resources can increase overall throughput.

​

Predicting failure risk

Expanse also predicts failure risk at submission time. This helps identify workloads that may be more likely to fail based on observed patterns.

Reducing failed workloads can improve cluster efficiency and free resources for productive jobs.

​

Learning from real cluster behavior

During the observation period, Expanse builds a model of how your cluster and its users actually operate.

This happens through passive telemetry gathered during normal cluster activity. No changes are required from cluster users during this time.

Over time, this analysis reveals:

• How workloads interact with cluster resources
• Where capacity is underutilized
• Opportunities to increase throughput