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:20260202T201804Z
LOCATION:266
DTSTART;TZID=America/Chicago:20251116T170500
DTEND;TZID=America/Chicago:20251116T172000
UID:submissions.supercomputing.org_SC25_sess221_ws_indis105@linklings.com
SUMMARY:Modular Architecture for High-Performance and Low Overhead Data Tr
 ansfers
DESCRIPTION:Rasman Mubtasim Swargo (Missouri University of Science and Tec
 hnology); Engin Arslan (Meta Plaforms, Inc.); and Md Arifuzzaman (Missouri
  University of Science and Technology)\n\nHigh-performance applications ne
 cessitate rapid and dependable transfer of massive datasets across geograp
 hically dispersed locations. Traditional file transfer tools often suffer 
 from resource underutilization and instability due to fixed configurations
  or monolithic optimization methods. We propose AutoMDT, a novel Modular D
 ata Transfer Architecture, to address these issues by employing deep reinf
 orcement learning based agent to simultaneously optimize concurrency level
 s for read, network, and write operations. This solution incorporates a li
 ghtweight network–system simulator, enabling offline training of a Proxima
 l Policy Optimization (PPO) agent in approximately 45 minutes on average, 
 thereby overcoming the impracticality of lengthy online training in produc
 tion networks. AutoMDT’s modular design decouples I/O and network tasks. T
 his allows the agent to capture complex buffer dynamics precisely and to a
 dapt quickly to changing system and network conditions. Evaluations on pro
 duction-grade testbeds show that AutoMDT achieves up to 8X faster converge
 nce and 68\% reduction in transfer completion times compared to state-of-t
 he-art solutions.\n\nRecording: Livestreamed, Recorded\n\nRegistration Cat
 egory: Technical Program Reg Pass, Workshop Reg Pass\n\nSession Chairs: Ez
 ra Kissel (Energy Sciences Network (ESnet)); Rafael Silva-Guimaraes (Feder
 al Institute of Espírito Santo, Department of Informatics); Nik Sultana (I
 llinois Institute of Technology, SCinet); Cees de Laat (University of Amst
 erdam, Lawrence Berkeley National Laboratory (LBNL)); Sonja Filiposka (Ss.
  Cyril and Methodius Univeristy in Skopje (UKIM) North Macedonia); and Akb
 ar Kara (Ciena Corporation, SCinet)\n\n
END:VEVENT
END:VCALENDAR