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:230
DTSTART;TZID=America/Chicago:20251120T111500
DTEND;TZID=America/Chicago:20251120T113000
UID:submissions.supercomputing.org_SC25_sess534_drs109@linklings.com
SUMMARY:Taming the Beast of Dynamic Resource Management in HPC
DESCRIPTION:Dominik Huber (Technical University of Munich)\n\nDynamic reso
 urce management (DRM) enables the resources assigned to a job to be adjust
 ed during execution. From a system perspective, DRM adds flexibility to re
 source allocation and job scheduling, with the potential to improve utiliz
 ation, throughput, energy efficiency, and responsiveness. From an applicat
 ion perspective, it allows users to match resource requests to evolving ne
 eds, potentially reducing queue times and costs.\n\nDespite these benefits
  and a decade of research, DRM remains largely an academic concept in HPC 
 rather than a production feature. This is due to the need for coordinated 
 changes across the entire software stack—applications, programming models,
  process managers, and resource managers—along with a holistic co-design e
 ffort to develop new scheduling and optimization policies.\n\nWe present a
  novel, end-to-end approach to DRM in HPC, introducing generic design prin
 ciples for parallel programming models that integrate applications’ dynami
 c process management with the resource managers’ optimization capabilities
 . We apply these principles across the HPC stack, incorporating standards 
 such as MPI and PMIx, to create a fully dynamic environment supporting div
 erse applications. This is paired with a performance-aware scheduling stra
 tegy based on steepest-ascent optimization.\n\nExperiments on up to 100 no
 des show moderate overheads for application process reconfiguration while 
 delivering substantial gains in system throughput and average job turnarou
 nd time compared to static scheduling under high-load conditions.\n\nTag: 
 Research & ACM SRC Posters\n\nRecording: Livestreamed, Recorded\n\nRegistr
 ation Category: Technical Program Reg Pass\n\nSession Chairs: Yanfei Guo (
 Argonne National Laboratory (ANL)); Shirley Moore (University of Texas at 
 El Paso); Kento Sato (RIKEN Center for Computational Science (R-CCS)); Chr
 is Schlipalius (Pawsey Supercomputing Research Centre; Commonwealth Scient
 ific and Industrial Research Organisation (CSIRO), Australia); and Anja Ge
 rbes (Georg-August-Universität Göttingen)\n\n
END:VEVENT
END:VCALENDAR