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:20260522T150115Z
LOCATION:C141/143/149
DTSTART;TZID=America/Chicago:20181113T103000
DTEND;TZID=America/Chicago:20181113T110000
UID:submissions.supercomputing.org_SC18_sess203_pap262@linklings.com
SUMMARY:GPU Age-Aware Scheduling to Improve the Reliability of Leadership 
 Jobs on Titan
DESCRIPTION:Christopher Zimmer, Don Maxwell, Stephen McNally, Scott Atchle
 y, and Sudharshan S. Vazhkudai (Oak Ridge National Laboratory)\n\nThe incr
 easing rate of failures on the Oak Ridge Leadership Computing Facility's (
 OLCF) Titan supercomputer, resulted in the replacement of 50% of its GPUs 
 between 2015 and 2017. The largest jobs, also known as "leadership jobs'',
  continued to experience increased application failures. These jobs contai
 ned significant amounts of low-failure rate and high-failure rate GPUs. Th
 e impacts of these failures were felt more by leadership jobs due to longe
 r wait times, runtimes, and higher charge rates. In this work, we have des
 igned techniques to increase the use of low-failure GPUs in leadership job
 s through targeted resource allocation. This employed two complementary te
 chniques, updating both the system ordering and the allocation mechanisms.
  In simulation, the application of these techniques resulted in a 33% incr
 ease in low-failure GPU hours being assigned to leadership jobs. Our GPU A
 ge-Aware Scheduling has been used in production on Titan since July of 201
 7.\n\nTag: GPUs, Resiliency, State of the Practice, System Software\n\nReg
 istration Category: Tech Program Reg Pass\n\nFinalist: BSP Finalist\n\nSes
 sion Chair: Ron Brightwell (Sandia National Laboratories)\n\n
END:VEVENT
END:VCALENDAR