Dynamic Partitions + Aggregation Designs. Est rows. How to handle/does it matter?


  • Hi all.

    I've created a 'template' partition with an aggregation design in my measure group (the slicer criteria is where OrderDate/100 = null so as to have 0 rows in the partition).

    I use the create XMLA from this partition (inc specifying the aggregation design) to dynamically create a new partition using SSIS based on the yyyymm of the incoming facts. This all works fine.

    I have some reservations about the metadata that i'm passing as part of the create partition xmla. Do I need to specify accurate Estimated Rows? Are these actually used by the engine similar to the distribution statistics on the db engine?

    And what about the est rows in the aggregation design? I cross joined factinternetsales in adventureworks a few times to get 200mil rows to test my dynamic partition creation ETL. But when i look @ the est rows in the agg design it shows 156 (this wouldve been the orig amount when i built it using the wizard). Does this matter? or do i need to maintain the full rowcount in the aggregation design specification as well? And how do I do it?

    Jakub @ Adelaide, Australia

    Thursday, November 14, 2013 1:24 AM

All replies