BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210916T132447Z
LOCATION:Ernesto Bertarelli
DTSTART;TZID=Europe/Stockholm:20210709T123000
DTEND;TZID=Europe/Stockholm:20210709T130000
UID:submissions.pasc-conference.org_PASC21_sess189_pap134@linklings.com
SUMMARY:Predictive, Reactive and Replication-based Load Balancing of Tasks
  in Chameleon and Sam(oa)²
DESCRIPTION:Paper\n\nPredictive, Reactive and Replication-based Load Balan
 cing of Tasks in Chameleon and Sam(oa)²\n\nSamfass, Klinkenberg, Chung, Ba
 der\n\nIncreasingly complex hardware architectures as well as numerical al
 gorithms make balancing load in parallel numerical software for adaptive m
 esh refinement an inherently difficult task, especially if variability of 
 system components and unpredictability of execution time comes into play. 
 Yet, traditional predictive load balancing strategies are largely based on
  cost models that aim to predict the execution time of computational tasks
 . To address this fundamental weakness, we present a novel reactive load b
 alancing approach in distributed memory for MPI+OpenMP parallel applicatio
 ns that is based on keepin tasks speculatively replicated on multiple MPI 
 processes. Replicated tasks are scheduled fully reactively without the nee
 d of a predictive cost model. Task cancelation mechanisms help to keep the
  overhead of replication minimal by avoiding redundant computation of repl
 icated tasks. We implemented our approach in the Chameleon library for rea
 ctive load balancing building upon previous work on reactive task migratio
 n. Our experiments in the parallel dynamic adaptive mesh refinement softwa
 re sam(oa)² demonstrate performance improvements in the presence
  of wrong cost models and artificially introduced noise to simulate imbala
 nces coming from hardware variability.\n\nDomain: CS and Math
END:VEVENT
END:VCALENDAR