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:20260422T143012Z
LOCATION:B302-B305
DTSTART;TZID=America/New_York:20241120T100000
DTEND;TZID=America/New_York:20241120T170000
UID:submissions.supercomputing.org_SC24_sess533_drs114@linklings.com
SUMMARY:Accelerating HPC Workflow Results and Performance Reproducibility 
 Analytics
DESCRIPTION:Kevin Assogba (Rochester Institute of Technology)\n\nModern hi
 gh-performance computing (HPC) workflows produce massive datasets, often e
 xceeding 100+ TB per day, driven by instruments collecting data at gigabyt
 es per second. These workflows, executed on advanced HPC systems with hete
 rogeneous storage devices, high-performance microprocessors, accelerators,
  and interconnects, are increasingly complex and often involve non-determi
 nistic computations. In this context, thousands of processes share computi
 ng resources using synchronization for consistency. The intricate process 
 interaction and existing non-deterministic operations challenge exploratio
 ns of workflow behaviors to ensure reproducibility, optimize performance, 
 and reason about what happens when processes compete for resources. Existi
 ng reproducibility analysis frameworks are not well-suited to identify the
  sources and locations of non-determinism and performance variations, as t
 hey often focus on the final workflow results and general statistics about
  workflow performance. \n\nWe address these challenges by introducing scal
 able techniques that accelerate intermediate workflow results' comparison 
 using variation-tolerant hashing of floating-point datasets, thus improvin
 g result reproducibility. We also capture workflow performance profiles an
 d benchmark various queries to analyze workflow performance reproducibilit
 y. We also identify opportunities to optimize the loading process and inde
 xing of performance data to ensure minimal initialization and querying ove
 rhead. Using collected performance data, we propose a cache-aware staggeri
 ng technique that leverages workflow I/O profiles to reduce bottlenecks an
 d resource contention, particularly in workflows that share the same input
  data. Our evaluations across molecular dynamics, cosmology, and deep lear
 ning workflows demonstrate significant speedup in intermediate results rep
 roducibility analyses compared to state-of-art baselines and our ability t
 o propose workflow execution strategies that maximize cache reuse and mini
 mize execution makespan.\n\nRegistration Category: Tech Program Reg Pass, 
 Exhibits Reg Pass\n\nSession Chairs: Ayesha Afzal (Friedrich-Alexander-Uni
 versität Erlangen-Nürnberg, Erlangen National High Performance Computing C
 enter); Sally Ellingson (University of Kentucky); and Alan Sussman (Univer
 sity of Maryland)\n\n
END:VEVENT
END:VCALENDAR