As supercomputers grow in size along with the number of users who access them, it becomes even more important to understand how your applications are performing.
Altair Mistral, previously Ellexus Mistral, is a lightweight application monitoring tool for HPC and scientific compute. Running all the time on a cluster, it provides live system telemetry and I/O monitoring.
Using Mistral, you can protect shared storage from rogue jobs and noisy neighbour applications through per-job I/O monitoring. You can catch problems before they happen with workflow qualification at scale and monitor system health with telemetry from every node and application.
This video demonstrates how to run Mistral to profile application I/O across jobs in a PBS Professional cluster. Viewers will learn how to set up a Mistral queue on PBS Professional using a job hook as well as the skills needed to launch Mistral using other schedulers. The demo visualises the results using Elastic Search and Grafana.