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:20210402T160046Z
LOCATION:Track 5
DTSTART;TZID=America/New_York:20201117T103000
DTEND;TZID=America/New_York:20201117T110000
UID:submissions.supercomputing.org_SC20_sess169_pap175@linklings.com
SUMMARY:Taming I/O Variation on QoS-Less HPC Storage: What Can Application
 s Do?
DESCRIPTION:Paper\n\nTaming I/O Variation on QoS-Less HPC Storage: What Ca
 n Applications Do?\n\nQiao, Liu, Podhorszki, Klasky, Chen\n\nAs high-perfo
 rmance computing (HPC) is being scaled up to exascale to accommodate new m
 odeling and simulation needs, I/O has continued to be a major bottleneck i
 n end-to-end scientific processes. This work aims to take advantage of the
  storage characteristics and explore application level solutions that are 
 interference-aware. In particular, we monitor the performance of data anal
 ytics and estimate the state of shared storage resources using discrete fo
 urier transform. If there is heavy I/O interference, data analytics can dy
 namically adapt to the environment by lowering the accuracy and performing
  partial or no augmentation from the shared storage, dictated by an augmen
 tation-bandwidth plot. We evaluate three data analytics; XGC, GenASiS and 
 Jet; on Chameleon, and quantitatively demonstrate that both the average an
 d variation of I/O performance can be vastly improved, with the mean and v
 ariance improved by as much as 18x and 60x, respectively, while maintainin
 g acceptable outcome of data analysis.\n\nTag: Algorithms, File Systems an
 d I/O, Storage\n\nRegistration Category: Tech Program Reg Pass
END:VEVENT
END:VCALENDAR