It was recently pointed out to me (by Hugo Lopes and Paulo Faria of Enabler in Portugal, who I had the pleasure of working with recently) that setting the ProcessingGroup property of a dimension can have quite a big impact on the amount of time it takes to process a dimension. By default this property is set to By Attribute, which means that each attribute in the dimension is processed separately using its own SQL Select statement which returns the distinct values for each member on the attribute. Here’s an example query used to process an attribute based on the Adventure Works Product dimension:
[dbo_DimProduct].[Color] AS [dbo_DimProductColor0_0]
FROM [dbo].[DimProduct] AS [dbo_DimProduct]
If you set it to By Table, however, in some circumstances (when your dimension is built from a single table?) only one SQL Select statement per table in the dimension structure is issued, reading all the data in, with AS presumably finding the distinct values itself.
When do you use which setting? I suppose it depends on a lot of things (performance of your data source, size and structure of your dimension, whether you’re processing other objects at the same time) so it’s going to be one of those things you’ll have to experiment with. Obviously most dimensions process extremely quickly anyway, but if you’ve got a large dimension which you need to process as quickly as possible this might come in handy.