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:20260522T150116Z
LOCATION:D163
DTSTART;TZID=America/Chicago:20181112T153000
DTEND;TZID=America/Chicago:20181112T155500
UID:submissions.supercomputing.org_SC18_sess142_ws_pdsw111@linklings.com
SUMMARY:Characterizing Deep-Learning I/O Workloads in TensorFlow
DESCRIPTION:Wei Der Chien, Stefano Markidis, and Chaitanya Prasad Sishtla 
 (KTH Royal Institute of Technology); Luis Santos (Institute Superior Técni
 co); Pawel Herman (KTH Royal Institute of Technology); Sai Narasimhamurthy
  (Seagate Systems UK); and Erwin Laure (KTH Royal Institute of Technology)
 \n\nThe performance of Deep-Learning (DL) computing frameworks rely on the
  performance of data ingestion and checkpointing. In fact, during the trai
 ning, a considerable high number of relatively small files are first loade
 d and pre-processed on CPUs and then moved to accelerator for computation.
  In addition, checkpointing and restart operations are carried out to allo
 w DL computing frameworks to restart quickly from a checkpoint. Because of
  this, I/O affects the performance of DL applications. In this work, we ch
 aracterize the I/O performance and scaling of TensorFlow, an open-source p
 rogramming framework developed by Google and specifically designed for sol
 ving DL problems. To measure TensorFlow I/O performance, we first design a
  micro-benchmark to measure TensorFlow reads, and then use a TensorFlow mi
 ni-application based on AlexNet to measure the performance cost of I/O and
  checkpointing in TensorFlow. To improve the checkpointing performance, we
  design and implement a burst buffer.  We find that increasing the number 
 of threads increases TensorFlow bandwidth by a maximum of 2.3× and 7.8× on
  our benchmark environments. The use of the tensorFlow prefetcher results 
 in a complete overlap of computation on accelerator and input pipeline on 
 CPU eliminating the effective cost of I/O on the overall performance. The 
 use of a burst buffer to checkpoint to a fast small capacity storage and c
 opy asynchronously the checkpoints to a slower large capacity storage resu
 lted in a performance improvement of 2.6× with respect to checkpointing di
 rectly to slower storage on our benchmark environment.\n\nTag: I/O, Storag
 e\n\nRegistration Category: Workshop Reg Pass\n\n
END:VEVENT
END:VCALENDAR