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:20210402T160544Z
LOCATION:Poster Module
DTSTART;TZID=America/New_York:20201119T083000
DTEND;TZID=America/New_York:20201119T170000
UID:submissions.supercomputing.org_SC20_sess337_rpost142@linklings.com
SUMMARY:Understanding I/O behavior of Scientific Deep Learning Application
 s in HPC systems
DESCRIPTION:Posters, Research Posters\n\nUnderstanding I/O behavior of Sci
 entific Deep Learning Applications in HPC systems\n\nDevarajan, Zheng, Sun
 , Vishwanath\n\nDeep learning has been widely utilized in various science 
 domains to achieve unprecedented results. These applications typically rel
 y on massive datasets to train the networks. As the size of datasets grow 
 rapidly, I/O becomes a major bottleneck in large scale distributed trainin
 g. We characterize the I/O behaviors of several scientific deep learning a
 pplications running on our production machine, Theta, at Argonne Leadershi
 p Computing Facility, with a goal to identify potential bottlenecks and to
  provide guidance for developing efficient parallel I/O library for scient
 ific deep learning. We found that workloads utilizing TensorFlow Data Pipe
 line can achieve efficient I/O through overlapping I/O with computation; h
 owever, they have potential scaling issues at larger scale as POSIX I/O is
  used underneath without parallel I/O.  We also identified directions for 
 I/O optimization for workloads utilizing a custom data streaming function.
  These workloads can potentially benefit from data prefetching, data sievi
 ng and asynchronous I/O.\n\nRegistration Category: Tech Program Reg Pass, 
 Exhibits Reg Pass
END:VEVENT
END:VCALENDAR