BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201805Z
LOCATION:264
DTSTART;TZID=America/Chicago:20251116T160000
DTEND;TZID=America/Chicago:20251116T161500
UID:submissions.supercomputing.org_SC25_sess204_ws_ss107@linklings.com
SUMMARY:Energy-Aware HPC Scheduling with LLM-Based Power Prediction
DESCRIPTION:Kevin Menear (National Renewable Energy Laboratory (NREL)); Al
 ex Wilkinson (University of Warwick); Tim Dykes and Utz-Uwe Haus (HPE, EME
 A Research Lab); and Dmitry Duplyakin (National Renewable Energy Laborator
 y (NREL))\n\nAs the increasing energy consumption of High-Performance Comp
 uting (HPC) systems places greater strain on electric grid infrastructure,
  operational strategies for load balancing become critically important. En
 ergy-aware scheduling offers a promising solution by enabling HPC systems 
 to function as actively managed loads within the energy grid. Despite exte
 nsive theoretical research on this strategy, practical implementations and
  real-system evaluations remain scarce. To bridge this gap, we introduce a
  systematic approach to developing, evaluating, and implementing energy-aw
 are scheduling without modifications to Slurm's core scheduler. Our method
  includes a novel mechanism for per-job power prediction based on Large La
 nguage Model embeddings of enriched job scripts, coupled with a lightweigh
 t, deployable scheduling strategy. Our predictor reduces per-job power MAE
  by 15% compared to the current state-of-the-art, and our simulated schedu
 ler shifts 4.0 MWh onto on-site solar without throughput loss. These resul
 ts demonstrate a clear and practical pathway to production deployment of e
 nergy-aware scheduling in HPC.\n\nRecording: Livestreamed, Recorded\n\nReg
 istration Category: Technical Program Reg Pass, Workshop Reg Pass\n\nSessi
 on Chairs: Mike Woodacre (Hewlett Packard Enterprise (HPE)); Michèle Weila
 nd (EPCC, The University of Edinburgh; The University of Edinburgh); Fumiy
 oshi Shoji (RIKEN Center for Computational Science (R-CCS), Center for Com
 putational Science); Pekka Manninen (CSC - IT Center for Science; Universi
 ty of Helsinki, Finland); James H. Rogers (Oak Ridge National Laboratory (
 ORNL)); and Cate Berard (US Department of Energy)\n\n
END:VEVENT
END:VCALENDAR