Uploaded image for project: 'SAFe Program'
  1. SAFe Program
  2. SP-1099

Standardize I/O performance benchmark

Details

    • Enabler
    • Not Assigned
    • PI7
    • None
    • Services
      • Benchmark is easy to discover (platform repo);
      • benchmark is easy to run (everything is scripted);
      • different benchmarks results (across teams, time and hardware platforms) can be compared meaningfully.
    • Hide
      • Benchmark is scripted end-to-end (from resource reservation to plots);
      • benchmark runs on at least to platforms (e.g. g5k and P3) with performance report for those platforms;
      • key metrics have been identified and documented;
      • scripts implement key metrics.
      Show
      Benchmark is scripted end-to-end (from resource reservation to plots); benchmark runs on at least to platforms (e.g. g5k and P3) with performance report for those platforms; key metrics have been identified and documented; scripts implement key metrics.
    • 2
    • 2
    • 0
    • Team_PLANET
    • Sprint 4
    • Hide

      Scripts have been extended and the latest changes have been pushed to the platform team's repo (https://gitlab.com/ska-telescope/platform-scripts/-/tree/master/iotest)
      Scripts have been documented, confluence page has been updated.
      Script has been demoed.

      Show
      Scripts have been extended and the latest changes have been pushed to the platform team's repo ( https://gitlab.com/ska-telescope/platform-scripts/-/tree/master/iotest ) Scripts have been documented, confluence page has been updated. Script has been demoed.
    • 8.2
    • Outcomes Reviewed, Demonstrated, Satisfies Acceptance Criteria, Accepted by FO
    • PI22 - UNCOVERED

    • Team_PLANET

    Description

      The I/O performance test could be valuable as a tool to evaluate potential hardware for SKA (e.g. for vendors) but 1/ further work is required to make it reproducible, and 2/ specific metrics need to be set in stone.

      Scripting work has been done in PI6 but it was tested on only one platform (Grid'5000). Adapting the scripts to at least another platform is important to identify and fix platform-dependent features.

      Scripts also need to be expanded to gather and plot more metrics (currently just data transfer speed as reported by the test). Reproducing the plots from memo 102 could be a sensible goal for this if nothing better has been identified.

      Finally, key specific metrics need to be identified and set in stone in the scripts, so future runs of the benchmark can be compared.

      Attachments

        Issue Links

          Structure

            Activity

              People

                b.mort Mort, Ben
                V.Lanore Lanore, Vincent [X] (Inactive)
                Votes:
                0 Vote for this issue
                Watchers:
                1 Start watching this issue

                Feature Progress

                  Story Point Burn-up: (100.00%)

                  Feature Estimate: 2.0

                  IssuesStory Points
                  To Do00.0
                  In Progress   00.0
                  Complete25.0
                  Total25.0

                  Dates

                    Created:
                    Updated:
                    Resolved:

                    Structure Helper Panel