BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163603Z
LOCATION:502-503-504
DTSTART;TZID=America/Denver:20191117T121000
DTEND;TZID=America/Denver:20191117T123000
UID:submissions.supercomputing.org_SC19_sess101_ws_dls106@linklings.com
SUMMARY:Aggregating Local Storage for Scalable Deep Learning I/O
DESCRIPTION:Workshop\n\nAggregating Local Storage for Scalable Deep Learni
 ng I/O\n\nZhang, Huang, Pauloski, Foster\n\nDeep learning applications int
 roduce heavy I/O loads on computer systems. The inherently long-running, h
 ighly concurrent, and random file accesses can easily saturate traditional
  shared file systems and negatively impact other users. We investigate her
 e a solution to these problems based on leveraging local storage and the i
 nterconnect to serve training datasets at scale. We present FanStore, a us
 er-level transient object store that provides low-latency and scalable POS
 IX-compliant file access by integrating the function interception techniqu
 e and various metadata/data placement strategies. On a single node, FanSto
 re provides performance similar to that of the XFS journaling file system.
  On many nodes, our experiments with real applications show that FanStore 
 achieves over 90% scaling efficiency.\n\nTag: Workshop Reg Pass, Deep Lear
 ning, Scientific Computing\n\nRegistration Category: Workshop Reg Pass, De
 ep Learning, Scientific Computing
URL:https://sc19.supercomputing.org/presentation/?id=ws_dls106&sess=sess10
 1
END:VEVENT
END:VCALENDAR