HPC systems often have more servers, CPU cores or GPU devices available than software entitlements allow, resulting in a need for “license aware” job handling to prevent job failures due to “license unavailable” errors. In this webinar, Chris Dagdigian, BioTeam’s Co-Founder and Senior Technical Director of Infrastructure, demonstrated how to integrate the popular commercial Schrödinger computational chemistry platform with AWS ParallelCluster HPC clusters running the Slurm batch scheduler including the specific steps needed to enable license-license aware job handling and scheduling for Schrödinger jobs and tasks.
The integration allows Schrödinger users working on the command-line or via the Maestro GUI to directly submit work to Parallelcluster HPC grids without fear of license-related job failures. Relative to the heavy open source software mix commonly seen in life science informatics, the computational chemistry, structural biology and molecular dynamics fields often involve usage of professionally developed, highly engineered and commercially licensed software. It is common for these commercial products to be managed or metered by license servers that control feature usage or the number of concurrent users allowed at any given time.
License tokens can be scarce and expensive resources and they are often consumed across a global organization including laptops, premise systems, and cloud servers—meaning that the availability of license entitlements can vary dynamically from moment to moment.