locked
Process update and attribute relationships RRS feed

  • Question

  • I understand that if you perform a process update and have flexible attribute relationships, that the process update drops all aggregations and bitmap indexes and these need to be processed again either with a process full of the measure group or process index. With ridgid relationships this does not happen.

    Are ridgid relationships really used that much? In all documentation on this people give the date dimension as an example, which for me isn’t that useful. I have a date dimension until 2020 so don’t need to worry about processing for many more years.

    If I think of common dimensions, product, store, customer, account etc. almost all attributes (apart from codes, ids etc) have the possibility to change, therefore have to be flexible right? And for me these type of dimensions are the dimensions that are frequently changing values and need to be updated via process update.

    Almost all of our dimensions have flexible relationships because the possibility of updates of attributes.

    This is starting to become a problem for us as a process update of dimensions and process full of selected measure groups is starting to become longer that a process full of the DB.

    We have found that the bulk of the time is during the dimensions process update dropping of the bitmap indexes due to flexible relationships of the dimension.

    Is there anything I’m missing or is a process full going to be the way forward?

    Thanks in advance for any help.


    • Edited by kiwiNspain Wednesday, June 24, 2015 2:47 PM
    Wednesday, June 24, 2015 2:45 PM

All replies

  • Hi,

    keep the count of metadata as small as possible. Where possible merge partitions. Eg we had an cube with over 3000  partitions and the bottleneck for processing was the step with dropping indices of the partitions. as a result also the commit phase was very long ~25 minutes which means NO new connections was possible to the whole instance.

    Wednesday, June 24, 2015 3:37 PM
  • Hi,

    for issue with longer processing time of process update  please refer this discussion,

    https://social.msdn.microsoft.com/Forums/sqlserver/en-US/1387f88a-81cc-4998-9990-c6ee77e9b0ac/processadd-longer-than-processfull?forum=sqlanalysisservices#b26cd454-b81e-4e28-8cbf-d5d12f5af3c1


    for optimization of your processing  use parallel processing

    use refe-How to Process a SSAS MOLAP cube as fast as possible

    URL-http://henkvandervalk.com/how-to-process-a-ssas-molap-cube-as-fast-as-possible-part-1

    Regards,

    Manish

    Thursday, June 25, 2015 5:50 AM