Forum Stats

  • 3,839,770 Users
  • 2,262,532 Discussions
  • 7,901,053 Comments

Discussions

EPM 11.2.6 HFM Performance Bug

1234579

Answers

  • Megan Michnick
    Megan Michnick Member Posts: 10 Blue Ribbon

    Oracle applied the patch in our test env and still no improvement in consolidation times. They also tried different values in these two referenced settings and consolidation times got worse. Oracle contacted microsoft as they have identified CPU issue when running multiple consols at one time(this is where we see severe degradation in performance). Issue related to business rules(vbcript). Basically issue is with Microsoft and waiting for resolution still.

  • Thanos A
    Thanos A Consolidation System Manager Member Posts: 1,443 Silver Trophy

    Hi @Megan Michnick ,

    Thank you for the news.

    Has Oracle said which business rules impact performance?

    Thank you,

    Thanos

  • TJHuisman
    TJHuisman Member Posts: 14 Blue Ribbon

    Just wanted to share that we (Stork) applied the patch in our test environment and it improved consolidation times by 16% (with ConsolidationMultiThreadScheme = 2, which was already the current setting).

    We are running 11.2.7 and tested both before and after the patch 3 times, using EPM Maestro (a full year, CAWD, all structures).

    Additionally it indeed solved the AMP issue in IC matching reports.

    So all in all quite happy (though we of course hoped for more 😉).

  • EPMTek
    EPMTek Member Posts: 14 Bronze Badge

    Good to hear!

    Is the 16% with or without UseConsolidationStatusCache=1?

    Steve

  • Megan Michnick
    Megan Michnick Member Posts: 10 Blue Ribbon

    Hi Thanos - it was stated that our business rules and they way they are written are not the issue, i guess it is how the windows 2019 OS processes how the calculations run. I am hoping there is resolution soon but Oracle is waiting to hear from microsoft.

  • TJHuisman
    TJHuisman Member Posts: 14 Blue Ribbon

    Without. It's still on zero/0. I missed that one in this thread... I'll go through it again to see what it does.

    I'm a bit afraid to make changes to these settings (currently don't have a lot of time to test either).

  • EPMTek
    EPMTek Member Posts: 14 Bronze Badge

    Check article Doc ID 2512975.1 for info AND caveats - as there is a potential downside.

    Steve

    TJHuisman
  • Thanos A
    Thanos A Consolidation System Manager Member Posts: 1,443 Silver Trophy

    After 3 years of working with this issue, I think that there are 2 different problems.

    i) The first is the one that Steve @EPMTek mentioned and it is fixed with UseConsolidationStatusCache=1. If I understand well, Oracle changed the way that the status of the cells is managed when you have more than 1 app server and slowed down the consolidation.

    ii)The second is the one that Megan @Megan Michnick encountered in which a change that comes from MS security patches impacts the consolidation performance. In our case, when you run more than 2 consolidations in parallel the whole performance is going really down up to 70%. In 2020, a consolidation took 7minutes and in 2022, same consolidation takes 45minutes to complete. It does not seem to be related with Network or DB but on how the CPU threads are prioritise to execute tasks where the number 12 is the magic number for servers with more than 18 vCores.

    I hope that Oracle and MS will soon fix both issues because it is shame that a tool as powerful as HFM cannot consolidate fast enough.

    Cheers,

    Thanos

  • TJHuisman
    TJHuisman Member Posts: 14 Blue Ribbon

    Thanks Steve! I've gone through the potential downside, but it's not completely clear to me whether:

    Will the parent with status "OK" (but with children with "CN" status) still be correctly consolidated by HFM when a consolidation is started, or is it then skipped because of the "OK" status?

    Thus really requiring a periodic "force calculation" on all base entities before doing a consolidation to make sure everything is correctly consolidated.

    I don't care too much about the status of the parents, as long as a consolidation is still triggered/performed on those parents as we've scheduled enough consolidation taskflows during closing days to have the correct consolidated numbers (thank you EPM Maestro). If a "force calculate" would still be required, the time saved with the setting would be lost again.

    It says in the doc ID: Running a consolidation on Parent will resolve the status mismatch (but I'm not sure if you need to do the workaround first). As I'm not native English, I'm still a bit in doubt.

  • EPMTek
    EPMTek Member Posts: 14 Bronze Badge

    Good question...

    I'm a little unclear too (but luckily I only worry about the Technical Infrastructure side these days...) since I thought that a Consolidate All with Data would ignore status anyway...

    If as @Thanos A mentioned it's due to when you have more than one HFM Java Server in the mix that this has an effect, then I wonder if it's to do with how changed data is signalled between servers.

    Sorry I can't be more help here

    Steve