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:20210527T134946Z
LOCATION:Digital
DTSTART;TZID=Europe/Stockholm:20200623T182500
DTEND;TZID=Europe/Stockholm:20200623T185000
UID:isc_hpc_ISC High Performance 2020_sess341_pap179@linklings.com
SUMMARY:Reinit++: Evaluating the Performance of Global-Restart Recovery Me
 thods For MPI Fault Tolerance
DESCRIPTION:Research Paper\n\nReinit++: Evaluating the Performance of Glob
 al-Restart Recovery Methods For MPI Fault Tolerance\n\nGeorgakoudis, Guo, 
 Laguna\n\nScaling supercomputers comes with an increase in failure rates d
 ue to\nthe increasing number of hardware components. In standard practice,
 \napplications are made resilient through checkpointing data and\nrestarti
 ng execution after a failure occurs to resume from the latest\ncheckpoint.
  However, re-deploying an application incurs overhead by\ntearing down and
  re-instating execution, and possibly limiting\ncheckpointing retrieval fr
 om slow permanent storage.\n\nIn this paper we present , a new design and 
 implementation of\nthe Reinit approach for global-restart recovery, which 
 avoids\napplication re-deployment. We extensively evaluate\ncontrasted wit
 h the leading MPI fault-tolerance approach of\nULFM, implementing global-r
 estart recovery, and the typical practice of\nrestarting an application to
  derive new insight on performance.\nExperimentation with three different 
 HPC proxy applications made\nresilient to withstand process and node failu
 res shows that \nrecovers much faster than restarting, up to 6 times, or U
 LFM, up to\n3 times, and that it scales excellently as the number of MPI p
 rocesses\ngrows.\n\nTag: Pre-Recorded
URL:https://2020.isc-program.com/presentation/?id=pap179&sess=sess341
END:VEVENT
END:VCALENDAR

