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:20181221T160727Z
LOCATION:D163
DTSTART;TZID=America/Chicago:20181112T112000
DTEND;TZID=America/Chicago:20181112T114500
UID:submissions.supercomputing.org_SC18_sess142_ws_pdsw114@linklings.com
SUMMARY:Evaluation of HPC Application I/O on Object Storage Systems
DESCRIPTION:Workshop\nI/O, Storage, Workshop Reg Pass\n\nEvaluation of HPC
  Application I/O on Object Storage Systems\n\nLiu, Koziol, Butler, Fortner
 , Chaarawi...\n\nPOSIX-based parallel file systems provide strong consiste
 ncy semantics, which many modern HPC applications do not need and do not w
 ant.  Object store technologies avoid POSIX consistency and are designed t
 o be extremely scalable, for use in cloud computing and similar commercial
  environments.  In this work, we evaluate three object store systems: Inte
 l DAOS, Ceph RADOS, and Openstack Swift, and evaluate them with three HPC 
 applications: VPIC, H5Boss, and BDCATS.  We have developed virtual object 
 layer (VOL) plugins for HDF5 that can redirect the applications' HDF5 call
 s to the underlying object storage systems' APIs, with minimum application
  code change.  Through our evaluation, we found that object stores have be
 tter scalability in many cases than POSIX file systems, but are not optimi
 zed for common HPC use cases, such as collective I/O.  Understanding curre
 nt object store I/O details and limitations will enable us to better desig
 n object stores for future HPC systems.
URL:https://sc18.supercomputing.org/presentation/?id=ws_pdsw114&sess=sess1
 42
END:VEVENT
END:VCALENDAR