2 Replies Latest reply: Jun 13, 2013 12:56 AM by James.zhang RSS

    Consolidation performance

    993723
    993723 May 31, 2013 8:26 AM
      Dear Experts,

      I am aware of MaxNumDataRecordsInRAM and MaxDataCacheSizeinMB settings.
      Having bumped up those settings to upper values (40M and 6000 currently), we can see during busy times HFM application (Hsvdatasource.exe) can take 30 GB of 32 GB RAM in the application server. Performance tuning guide suggests we could even increase more like 60M and 9000 if we have 32 GB of RAM given that only memory intensive application is HFM as in our case.

      can you please share your experience for those parameters ? is it possible to improve consolidation speed by focusing more on those parameters or else ?

      Best Regards,
      • 407 Views
      • Tags:
        • 1. Re: Consolidation performance
          neels.grobler
          neels.grobler Jun 11, 2013 8:02 AM (in response to 993723)

          Not sure how many users you have working on your system and how much data you are consolidating, but 30 GB sounds like a lot for an HFM App. You can drastically affect speed depending on an App's rules, and in my experience, you can get a much bigger performance improvement by improving the rules than by optimizing those settings.

          • 2. Re: Consolidation performance
            James.zhang
            James.zhang Jun 13, 2013 12:56 AM (in response to 993723)

            Hi,

             

            I am afraid you have data explosion at HFM which is caused by defects in rule file.

             

            For example, if any rules writes like

             

            Hs.Exp A#10000 = A#120000.I#[ICP Top].C1#ABC.C2#ABC.C3#ABC.C4#ABC

             

            If A#120000.I#[ICP Top].C1#ABC.C2#ABC.C3#ABC.C4#ABC = 100, then this 100 will copied to all the available intersection for account 10000.

             

            However, you may only need to copy this value to one intersection for account 10000.

             

            This is the most possible reason that you have so many data at HFM.

             

            Thanks.

             

            James Zhang