The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Abstract: To maintain a robust and reliable supercomputing hardware system there is a critical need to understand various system events, including failures occurring in the system. Toward this goal, we analyze various system logs such as error logs, job logs and environment logs from Argonne Leadership Computing Facility's (ALCF) Theta Cray XC40 supercomputer. This log data incorporates multiple subsystem and component measurements at various fidelity levels and temporal resolutions - a very diverse and massive dataset. To effectively identify various patterns that characterize system behavior and faults over time, we have developed a visual analytics tool, MELA, to better identify patterns and glean insights from these log data.