Workshop: Thirteenth SC Workshop on Best Practices for HPC Training and Education
Authors: Smita Darmora (University of Illinois at Chicago) and Jeffrey Weekley (University of California, Santa Cruz)
Abstract: High-performance computing training reliably teaches users how to submit jobs. It rarely equips them to reason about why their work behaves as it does on a shared system. At a PEARC26 Birds-of-a-Feather session on system literacy, practitioners nearly all agreed that users should hold a working mental model of system behavior, and nearly all reported that users do not. We argue that this gap persists because system literacy is formed incidentally rather than taught: users assemble their own models from documentation, experience, and chance interactions, producing models that vary widely in accuracy and surface as uneven productivity on shared infrastructure. We locate the mechanism by which such models are revised in productive struggle, the calibrated difficulty skilled facilitators already sustain when they diagnose the misconception behind an error rather than only correcting it. Whether that practice can be scaled is the question we pose.
Back to Thirteenth SC Workshop on Best Practices for HPC Training and Education Archive Listing
Back to Full Workshop Archive Listing