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:20260202T201804Z
LOCATION:264
DTSTART;TZID=America/Chicago:20251116T143000
DTEND;TZID=America/Chicago:20251116T144500
UID:submissions.supercomputing.org_SC25_sess204_ws_ss101@linklings.com
SUMMARY:HPC Digital Twins for Evaluating Scheduling Policies, Incentive St
 ructures and their Impact on Power and Cooling
DESCRIPTION:Matthias Maiterth and Wesley H. Brewer (Oak Ridge National Lab
 oratory (ORNL)); Jaya S. Kuruvella, Arunavo Dey, and Tanzima Z. Islam (Tex
 as State University); Rashadul Kabir (Colorado State University); Kevin Me
 near and Dmitry Duplyakin (National Renewable Energy Laboratory); Tapasya 
 Patki (Lawrence Livermore National Laboratory (LLNL)); and Terry Jones and
  Feiyi Wang (Oak Ridge National Laboratory (ORNL))\n\nSchedulers are criti
 cal for optimal resource utilization in high-performance computing. Tradit
 ional methods to evaluate schedulers are limited to post-deployment analys
 is, or simulators, which do not model associated infrastructure. In this w
 ork, we present the first-of-its-kind integration of scheduling and digita
 l twins in HPC. This enables what-if studies to understand the impact of p
 arameter configurations and scheduling decisions on the physical assets, e
 ven before deployment, or regarching changes not easily realizable in prod
 uction. We (1) provide the first digital twin framework extended with sche
 duling capabilities, (2) integrate various top-tier HPC systems given thei
 r publicly available datasets, (3) implement extensions to integrate exter
 nal scheduling simulators. Finally, we show how to (4) implement and evalu
 ate incentive structures, as-well-as (5) evaluate machine learning based s
 cheduling, in such novel digital-twin based meta-framework to prototype sc
 heduling. Our work enables what-if scenarios of HPC systems to evaluate su
 stainability, and the impact on the simulated system.\n\nRecording: Livest
 reamed, Recorded\n\nRegistration Category: Technical Program Reg Pass, Wor
 kshop Reg Pass\n\nSession Chairs: Mike Woodacre (Hewlett Packard Enterpris
 e (HPE)); Michèle Weiland (EPCC, The University of Edinburgh; The Universi
 ty of Edinburgh); Fumiyoshi Shoji (RIKEN Center for Computational Science 
 (R-CCS), Center for Computational Science); Pekka Manninen (CSC - IT Cente
 r for Science; University of Helsinki, Finland); James H. Rogers (Oak Ridg
 e National Laboratory (ORNL)); and Cate Berard (US Department of Energy)\n
 \n
END:VEVENT
END:VCALENDAR