Announcement

Collapse
No announcement yet.

Need help slicing and dicing test results

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Need help slicing and dicing test results

    Ran fio test with lots of options: 4 workload choices, 2 buffered choices, 2 direct choices, 2 sync choices, and way too many block size choices: https://openbenchmarking.org/result/...IB-FIOEVERYT76

    The data is all there, but we can't get the results page to display the right graphs. It makes no sense to compare different workloads, so that should be constant in any graph. Then, since there are so many options we'll freeze several of them for each graph and compare the remaining one or two variables. So, what we want is:
    Comparison of different block sizes for unbuffered direct sync sequential write
    Comparison of direct vs indirect for buffered sync 4K blocks sequential read
    Comparison of buffered vs unbuffered for direct sync 4M blocks random write
    &c.

    Then once we determine optimal settings for this particular device (7200rpm HDD in USB2 attached SATA enclosure), we have more results for the same variables on different devices: 7200rpm HDD USB3 attached SATA enclosure
    7200rpm SAS HDD on PCIEx2 SAS controller
    weird variable rpm SATA HDD on PCIEx2 SAS controller

    Is that kind of result mangling possible in the result viewer? Or anything else? Can results be converted to CSV and then mangled in Libreoffice?

    Thanks!

  • #2
    Click the " Condense Multi-Option Tests Into Single Result Graphs" option in the result viewer. e.g. makes your URL as https://openbenchmarking.org/result/...EVERYT76&cts=1

    That will show possible different comparisons.

    Or otherwise for exporting to CSV it's right there in OB or the result viewer from the upper right -> Export Benchmark Data.
    Michael Larabel
    https://www.michaellarabel.com/

    Comment


    • #3
      That didn't quite do what we were looking for when we tried before, but we'll try again. Should have spent more time researching before posting, sorry.

      Comment

      Working...
      X