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:20260202T201803Z
LOCATION:264
DTSTART;TZID=America/Chicago:20251116T164500
DTEND;TZID=America/Chicago:20251116T170000
UID:submissions.supercomputing.org_SC25_sess204_ws_ss113@linklings.com
SUMMARY:EMLIO: Minimizing I/O Latency and Energy Consumption for Large-Sca
 le AI Training
DESCRIPTION:Md Hasibul Jamil (University at Buffalo SUNY), MD S. Q. Zulkar
  Nine (Tennessee Technological University), and Tevfik Kosar (University a
 t Buffalo (SUNY))\n\nLarge-scale deep learning workloads increasingly face
  I/O bottlenecks as datasets exceed local storage and GPU compute outpaces
  network and disk speeds. While recent systems optimize data-loading time,
  they often ignore I/O energy costs—a critical factor at scale. We present
  EMLIO, an Efficient Machine Learning I/O service that minimizes both end-
 to-end data-loading latency (𝑇) and I/O energy consumption (𝐸) across vari
 able-latency networked storage. EMLIO uses a lightweight data-serving daem
 on on storage nodes to serialize and batch raw samples, stream them over T
 CP with out-of-order prefetching, and integrate with GPU-accelerated (NVID
 IA DALI) preprocessing on the client side. In evaluations over local disk,
  LAN (0.05 ms & 10 ms RTT), and WAN (30 ms RTT), EMLIO achieves up to 8.6×
  faster I/O and 10.9× lower energy use than state-of-the-art loaders, main
 taining constant performance and energy profiles across distances. Its ser
 vice-based architecture offers a scalable blueprint for energy-aware I/O i
 n next-generation AI clouds.\n\nRecording: Livestreamed, Recorded\n\nRegis
 tration Category: Technical Program Reg Pass, Workshop Reg Pass\n\nSession
  Chairs: Mike Woodacre (Hewlett Packard Enterprise (HPE)); Michèle Weiland
  (EPCC, The University of Edinburgh; The University of Edinburgh); Fumiyos
 hi Shoji (RIKEN Center for Computational Science (R-CCS), Center for Compu
 tational Science); Pekka Manninen (CSC - IT Center for Science; University
  of Helsinki, Finland); James H. Rogers (Oak Ridge National Laboratory (OR
 NL)); and Cate Berard (US Department of Energy)\n\n
END:VEVENT
END:VCALENDAR