Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-6785

Scale up eMRCf benchmark

    • Type: Icon: Task Task
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None

      In WT-6776 I presented results from an untuned "first time" run of my benchmark for testing WT performance with eMRCf=true. To get a more realistic picture of any performance overhead introduced here (beyond WT-6681) we scale up this basic benchmark to something that will stretch the system a bit more.

      Specifically we should:

      • Operate on a larger data set.  Increasing the date set 10x would give us a collection of 10 million documents probably consuming about 2.3 GB after the initial population stage.
      • Perform more operations during the benchmark stage.  Again a 10x increase in the number of operations would be a good first step.  We'll spread these operations over more documents as well, since we'll have a larger data set.  This will create more cache pressure.
      • Increase the number of threads used during the benchmark stage. Currently the benchmark is single threaded, because using a large number of client threads on my 2 CPU virtual workstation means that lots of parallel requests just creates lots of queuing delay.  Larger instances are available in Evergreen.  We should test using those.

            Assignee:
            keith.smith@mongodb.com Keith Smith
            Reporter:
            keith.smith@mongodb.com Keith Smith
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

              Created:
              Updated:
              Resolved: