BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160546Z
LOCATION:Track 10
DTSTART;TZID=America/New_York:20201113T175000
DTEND;TZID=America/New_York:20201113T175500
UID:submissions.supercomputing.org_SC20_sess232_pec241@linklings.com
SUMMARY:A Parallel Job Scheduling Method to Effectively Use Shared Heterog
 eneous Systems for Urgent Computations
DESCRIPTION:Workshop\n\nA Parallel Job Scheduling Method to Effectively Us
 e Shared Heterogeneous Systems for Urgent Computations\n\nAgung\n\nLightni
 ng talk: Dedicated resources are widely used in HPC systems for urgent dec
 ision-making because they can provide immediate and dedicated access to ur
 gent computations.  Setting up dedicated resources for each urgent computa
 tion, however, is not economically viable because the computations may occ
 ur rarely, and require huge amounts of resources.  Moreover, disasters can
  cause damage to dedicated resources that makes it impossible to perform u
 rgent computations that can help to mitigate the damage.  Due to these lim
 itations, using existing shared infrastructures becomes an indispensable a
 pproach for HPC systems supporting urgent decision making.\n\nRecent studi
 es had shown that providing job scheduling methods that can handle urgent 
 jobs is one important challenge that must be addressed to enable shared in
 frastructures for urgent computations.  Since the shared infrastructures a
 lso serve regular uses of the resource, considering both urgent and regula
 r jobs is necessary because both jobs can delay each other. Moreover, pree
 mpting regular jobs is necessary to guarantee immediate execution of urgen
 t jobs.  In heterogeneous systems with coprocessors, preemption becomes mo
 re challenging because it relies on system functionalities provided by the
  host processor.  In this ongoing work, we introduce a parallel job schedu
 ling method to effectively use shared heterogeneous systems for urgent com
 putations. To achieve low preemption delays, we employ an in-memory proces
 s swapping mechanism to preempt jobs running on coprocessors.  Our evaluat
 ion with workloads of real HPC systems shows that our method can achieve n
 egligible delays of urgent jobs without significant increases in response 
 times and slowdowns of regular jobs.\n\nTag: Applications, Scientific Comp
 uting, Simulation, State of the Practice, Technology Challenge\n\nRegistra
 tion Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR