Tuning Workflow Management System on Shared HPC Resources

Recruiting: mentee

Learn to tune WMS-based pipelines on shared HPC. For admins and RSEs: master the fairshare-and-backfill mechanics and the three tuning tradeoffs behind them. Monthly cohort, one semester. Bring your own pipeline and cluster—leave with a benchmarked, tuned config.

mentee preferred attributes:

HPC SysAdmins, RSEs, CI Facilitators, Researchers, and more are all welcome!

I'm Interested I have a question

Details

Format: 1 mentor : N mentees, monthly group discussion, starting from August - September 2026.
Content: what a WMS is and its pros/cons on HPC; how WMS jobs actually hit Slurm (native submission vs. job arrays vs. meta-schedulers); tuning knobs at the WMS, Slurm, and site level.
First ~3 session: official Nextflow tutorial + the HPC Carpentry
Symposium: Bring your own pipeline, your own cluster, your choice of WMS & literature.
By session 6: draft a site-specific Slurm tuning profile for WMS workloads (partition routing, submit caps, fairshare posture) for your own cluster; peer-reviewed by the group.
By session 8: benchmark one of your own WMS pipelines under at least two Slurm execution configurations (e.g., native submission vs. job array); produce a before/after report on walltime and scheduler overhead.
Deliveries:
- Each participant leaves with one tuned, benchmarked configuration running at home, and one peer-reviewed tuning profile.
- The cohort builds a shared, open resource of WMS-on-HPC patterns and anti-patterns that outlives the mentorship (potential paper)
- A standing cross-institution peer network for the next WMS problem that comes up
- Exposure & engagement with border WMS communities