HEPiX Benchmarking Working Group

Friday 21 Sept 2018, 13:55 → 15:20 Europe/Zurich

31/S-023 (CERN)

Manfred Alef (Karlsruhe Institute of Technology (KIT)), Domenico Giordano (CERN), Michele Michelotto

14:00 → 14:10 News

Speakers: Domenico Giordano (CERN), Manfred Alef (Karlsruhe Institute of Technology (KIT)), Michele Michelotto

1. CHEP 2018 proceedings

   • the deadline for the initial submission of your articles is: Wed Oct 31 at 23:59 UTC

   • The CHEP 2018 proceedings will be published in the EPJ Web of Conferences:

     https://www.epj-conferences.org/

     Most of the instructions for authors are provided on the following page:

     https://www.epj-conferences.org/for-authors

 

14:10 → 14:25 Benchmark studies and impact of Virtualization, HT and OS

Speaker: Benoit Roland (Albert Ludwigs Universitaet Freiburg (DE))

Benchmarking-roland-21-09.pdf

14:25 → 14:45 Summary of Benchmark Discussions at pre-GDB, GDB, and MB

A) Reminder

• Requirements for a benchmark suite
  • Must be usable not only by experts but also by site procurement teams and hardware vendors
    • Simple instructions: Insert disk, run shell script, wait, and read and report score – No remote data access from vendor‘s office
    • Clear license conditions
    • Long term commitment from the experiments
  • Provide the software support for several years
  • Tamper-proof repository, checksums, strict version control
• This is feasible with technologies already adopted (cvmfs, containers, ...) but does not come for free for the experiments

 

B) Summer updates:

• Recent development from EP-SFT (CernVM team) enables snapshotting of software from CVMFS. 
  • For more information Efficient unpacking of required software from CernVM-FS
  • Tested by developers on the KV benchmarking code 
  • Future work
    • To be now extended to the experiments' workloads documented in those pages CMS, ATLAS
    •  
• Manpower: A CERN IT Technical student will start in October her work on benchmarking.
  • First goal is to test the above packaging and report to the Experiments for further validation/testing.
  • Other contributions from the WG members is welcome.