Protecting shared storage from bad I/O patterns with HPCWorks Mistral

Enterprise high-performance computing (HPC) workload managers efficiently allocate jobs and orchestrate CPU, memory, license and GPU sharing, but they’re not designed to control workloads’ I/O patterns, which may be overlooked when tuning HPC environment performance. It’s easy for a job with bad I/O patterns to overload shared storage, often when a user tests a workflow on one or two nodes and then scales it without understanding how I/O patterns scale.

HPCWorks™ Mistral™ software is the leading application monitoring tool for HPC and scientific computing, with unrivaled ability to track I/O patterns across HPC clusters. Download the fact sheet to learn how it monitors I/O, CPU and memory to quickly locate rogue jobs and storage bottlenecks and keep track of what’s running day-to-day.