Dimensionality of facts and Microsoft SSAS Cubes

[sman]

Hello,

My understanding is that dimensionality of facts is important and careful attention has to paid when facts of different dimensionality are combined in analysing. So I would expect that facts are grouped in to measure groups (ie. having the same dimensionality) in Cubes so that end users know which facts can be combined / used in the same report, when analysing data by slicing and dicing.

However, I notice that MS Analysis Services Cubes do not make this distinction. All the facts (measures), regardless of their dimensionality, are in one common please called 'Measures Dimension'. There is no easy for a cube user to know the dimensionality of facts (measures). This paints the picture that dimensionality of facts has no importance when analysing/slicing/dicing Cube data. I thought this was a big "NO NO", but no one has found any problems with this.

Has anyone noticed this? Or, am I mistaken and dimensionality of facts is not important?

Thanks

[Mike Honey]

Hi Sman,
This may be a "feature" of your client tool.

If you query an SSAS cube using Excel 2007+ PivotTables, SSRS Report Designer or the BIDS cube browser, the Measure Groups (Fact Tables) defined in your cube are shown. Excel also lets you filter the Dimension list by Measure Group: "Show fields related to".

Good luck!

[ngalemmo]

I don't know, I may be out of touch, but doesn't a cube have only one set of dimensions? Some measures may not be fully additive, but you would not have measures of different grains in the same cube.

MS's implementation of MDX does support joining cubes, but only along common dimensions.

[hang]

Dimensionality is very important in any OLAP cube. SSAS is definitely no exception. Ignoring it will result in unexpected result by cube. However cube does not stop you from connecting fact to irrelevant dimension, it's users' responsibility to relate dimensions to the fact and configure dimension usage properly, based on the knowledge on the underlying schema.

What do you mean “Measures dimension”? As far as I know, there are clearly two tree-view boxes, Measures and Dimensions, that contain the two kinds separately. However there is a MS term “Fact dimension” which is effectively the degenerate dimension, meaning an attribute or measure can be configured as a dimension.

[Mike Honey]

@ngalemoo

Since SQL Server 2005, SSAS has allowed multiple facts ("Measure Groups") of varying grain in a single cube.

Mike

[Sponsored content]