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:C140/142
DTSTART;TZID=America/Chicago:20181113T163000
DTEND;TZID=America/Chicago:20181113T170000
UID:submissions.supercomputing.org_SC18_sess182_pap311@linklings.com
SUMMARY:Mitigating Inter-Job Interference Using Adaptive Flow-Aware Routin
 g
DESCRIPTION:Staci A. Smith, Clara E. Cromey, and David K. Lowenthal (Unive
 rsity of Arizona); Jens Domke (Tokyo Institute of Technology); and Nikhil 
 Jain, Jayaraman J. Thiagarajan, and Abhinav Bhatele (Lawrence Livermore Na
 tional Laboratory)\n\nOn most high performance computing platforms, applic
 ations share network resources with other jobs running concurrently on the
  system.  Inter-job network interference can have a significant impact on 
 the performance of communication-intensive applications, and no satisfacto
 ry solutions yet exist for mitigating this degradation.\n\nIn this paper, 
 we analyze network congestion caused by multi-job workloads on two product
 ion systems that use popular network topologies---fat-tree and dragonfly. 
 For each system, we establish a regression model to relate network hotspot
 s to application performance degradation, showing that current routing str
 ategies are insufficient to load-balance network traffic and mitigate inte
 rference on production systems.  We then propose an alternative type of ad
 aptive routing strategy, which we call adaptive flow-aware routing.  We im
 plement a prototype of our strategy, and tests on the fat-tree system show
  up to a 46% improvement in job run time when compared to the default rout
 ing.\n\nTag: Networks, Resource Management, Scheduling, State of the Pract
 ice, System Software\n\nRegistration Category: Tech Program Reg Pass\n\nFi
 nalist: BSP Finalist\n\nSession Chair: Bronis R. de Supinski (Lawrence Liv
 ermore National Laboratory (LLNL), Queens University Belfast (QUB))\n\n
END:VEVENT
END:VCALENDAR
