Planning Analytics

Planning Analytics

Get AI-infused integrated business planning

 View Only

  • 1.  Performance issues where Cube is the data source

    Posted 12/02/20 12:11 AM
    Thought I would post in case anyone else is having the same issue.

    I have some processes that roll forward values from one month to a range of target months e.g. roll forward prices, standard costs etc.
    My data source is a cube view and contains only N: level elements.

    As a standard, I use the following against the view:
    ViewExtractSkipZeroesSet(sCube,sView,1);
    ViewExtractSkipCalcsSet(sCube,sView,0);
    ViewExtractSkipRuleValuesSet(sCube,sView,0);

    The values I am reading from are not rule derived and like mentioned, all leaf level so no need to worry about setting anything other than the suppression of zeroes.

    Probably in the region of 3,000 cells are returned in the view so very tiny by all accounts.

    When I run the process on my cloud instance which is apparently running on 2.0.9.1, the process took 812 seconds to complete.
    If I run the same process on the same data set on a local instance running 2.0.9.2 (11.8.00100.13), the process completes in about a second.

    After some fiddling and some assistance, there are two tweaks that seem to alleviate this.
    1) Set ViewExtractSkipCalcsSet to 1, even if you only have N: levels
    2) Leave ViewExtractSkipCalcsSet as 0 but Incorporate C: levels in one dimension and skip these in Meta data/data as applicable.

    Hope this helps for those that may not be able to upgrade or those on cloud where the upgrade will happen at some point.

    ------------------------------
    George Tonkin
    ------------------------------

    #PlanningAnalyticswithWatson


  • 2.  RE: Performance issues where Cube is the data source

    Posted 12/07/20 10:14 AM
    I noticed a performance hit a while back as well (not quite sure about the correlation with product version), although not as severe as yours (more like 1 hour becomes 10 hours). However, I'm running the same 11.8.00100.13 build locally as you have good performance with. My solution was to have the process write the results into files and then load the data from those into the target cubes.

    ------------------------------
    Anssi Lehtinen
    ------------------------------



  • 3.  RE: Performance issues where Cube is the data source

    Posted 12/07/20 11:02 AM
    HI Anssi,

    Thanks for the feedback. Concerned that you are seeing this behaviour on the same version that seems to be working fine for me.
    Definitely going to avoid exporting to files and importing as there should be no reason that it should not just work.  I have ruled out rules and feeders by the way - removed these from the cube I am reading and writing to.

    Upgrading tonight so will hopefully be able to test and provide feedback tomorrow.

    ------------------------------
    George Tonkin
    ------------------------------

    Original Message


  • 4.  RE: Performance issues where Cube is the data source

    Posted 12/08/20 02:13 PM
    So after an upgrade from 2.0.9.1 base (11.8.00000.33), no IFs applied, to 2.0.9.3 (11.8.00200.24) the issue seems to have been resolved.

    Processes are back to a second or two to complete.

    ------------------------------
    George Tonkin
    ------------------------------

    Original Message


Related Content