Performance analysis and visualization framework for scientific workflows

Areas: Development tools, Uses AI

CASS member: RAPIDS

Description

Chimbuko is the first scalable, workflow-level performance analysis tool. In situ, real-time anomaly detection algorithms are applied to application trace data to isolate abnormal performance, thereby reducing the massive volume of trace data generated by large-scale HPC workflows to a manageable level while retaining information required to identify performance issues with a finer, event-level granularity than a broad-level profile.

Target audience

Domain scientists, HPC users/developers.

Additional resources