BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000712Z
LOCATION:607
DTSTART;TZID=America/Denver:20231113T110000
DTEND;TZID=America/Denver:20231113T110500
UID:submissions.supercomputing.org_SC23_sess448_ws_canolt110@linklings.com
SUMMARY:eBPF-Based Performance Fingerprint of Containerized HPC Applicatio
 ns
DESCRIPTION:Maximilian Hoeb (Ludwig-Maxmilians-Universität München, Leibni
 z Supercomputing Centre)\n\nHighly optimized systems face the challenge of
  meeting the computational demands of domain-specific simulations and work
 flow applications. Those range from low- to high-level implementations and
  differ widely regarding optimization and requirements, and affecting incr
 easingly the centers' energy efficiency. With today’s heterogeneous cluste
 rs, containerization allows arbitrary applications to bypass architectural
  differences, and focus mainly on core-functions instead of deployment iss
 ues.\n\nOur framework determines general performance characteristics of co
 ntainerized HPC-applications with unknown behavior, and satisfies computat
 ional, memory, and interconnect requirements independent of the container-
 technology. It allows developers and administrators to evaluate runtime pa
 rameters of blackbox applications, without inspecting the inside of any co
 ntainer based on kernel-level measurements with eBPF. We derived first alg
 orithms for the Container-Fingerprint, a quantified and comparable runtime
  characteristic, enabling optimized mapping to target systems.\n\nFor eval
 uation we investigate benchmark applications on different architectures ty
 pical for supercomputing centers. Measurements indicate that the derived f
 ingerprints are suitable to distinguish the performance of containers and 
 systems allowing an optimized allocation of HPC-containers. By applying th
 e scheme we work towards a twofold improvement: an increase in the efficie
 ncy of system usage and energy consumption, and a deployment optimization 
 of containers by enabling streamlined, requirements-oriented allocations o
 ptimizing the balance between resource-usage and time-to-solution.\n\nRegi
 stration Category: Workshop Reg Pass\n\nSession Chairs: Richard Shane Cano
 n (Lawrence Berkeley National Laboratory (LBNL)); Alberto Madonna (ETH Zür
 ich, Swiss National Supercomputing Centre (CSCS)); Laurie A. Stephey (Lawr
 ence Berkeley National Laboratory (LBNL), National Energy Research Scienti
 fic Computing Center (NERSC)); and Andrew Younge (Sandia National Laborato
 ries)\n\n
END:VEVENT
END:VCALENDAR