Why doesn't my dataflow work?! ...dataflows from n to (n+2) dimensions

Document created by Bob Gill on Jan 16, 2018
Version 1Show Document
  • View in full screen mode

Consider a situation like the following, where we want to copy data from one cube to another, but there are two additional dimensions in the target.

 

Source Cube DimensionsTarget Cube Dimensions
  • Month (D)
  • Product (D)
  • Month (D)
  • Product (S)
  • Geography (S)
  • Version (S)

 

When I first encountered this, I thought I could just add a SELECT step to specify the Geography and Version to which I want to save my data. Sometimes, that works. Sometimes, we need an intermediate cube to take one intermediate step. By copying the data to this intermediate cube first, we can continue to use the high performance mode. We can then copy from this intermediate cube to the target, again using high performance mode.

 

Source Cube DimensionsIntermediate Cube DimensionsTarget Cube Dimensions
  • Month (D)
  • Product (D)
  • Month (D)
  • Product (S)
  • Geography (S)
  • Month (D)
  • Product (S)
  • Geography (S)
  • Version (S)

 

There are other alternatives to solve this too. By making more of the target dimensions dense, we don't need an intermediate cube. On larger cubes, this may negatively impact performance when copying data.

 

Thanks Coulton Carrol for sharing this design approach with me.

2 people found this helpful

Attachments

    Outcomes