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:20260422T143141Z
LOCATION:B309
DTSTART;TZID=America/New_York:20241120T160000
DTEND;TZID=America/New_York:20241120T163000
UID:submissions.supercomputing.org_SC24_sess378_pap236@linklings.com
SUMMARY:Workflow Roofline Model for End-to-end Workflow Performance Analys
 is
DESCRIPTION:Nan Ding, Brian Austin, Yang Liu, Neil Mehta, Steven Farrell, 
 Johannes Blaschke, Leonid Oliker, Hai Ah Nam, Nicholas Wright, and Samuel 
 Williams (Lawrence Berkeley National Laboratory (LBNL))\n\nAs next-generat
 ion experimental and observational instruments for scientific research are
  being deployed with higher resolutions and faster data capture rates, the
  fundamental demands of producing high-quality scientific throughput requi
 re portability and performance to meet the high productivity goals. Unders
 tanding such a workflow's end-to-end performance on HPC systems is formida
 ble work. In this paper, we address this challenge by introducing a Workfl
 ow Roofline model, which ties a workflow's end-to-end performance with pea
 k node- and system- performance constraints. We analyze four workflows: LC
 LS, a time-sensitive workflow that is bound by system external bandwidth; 
 BerkeleyGW, a traditional HPC workflow that is bound by node-local perform
 ance; CosmoFlow, an AI workflow that is bound by the CPU preprocessing; an
 d GPTune, an auto tuner that is bound by the data control flow. We demonst
 rate the ability of our methodology to understand various aspects of perfo
 rmance and performance bottlenecks on workflows and systems and motivate w
 orkflow optimizations.\n\nTag: Heterogeneous Computing, Linear Algebra, Ne
 twork, Parallel Programming Methods, Models, Languages and Environments, P
 erformance Evaluation and/or Optimization Tools\n\nRegistration Category: 
 Tech Program Reg Pass\n\nAward Finalist: Best Paper Finalist\n\nSession Ch
 air: Sarah Neuwirth (Johannes Gutenberg University Mainz)\n\n
END:VEVENT
END:VCALENDAR