Number of Dimensions

[dssd]

How many dimensions are a good practice in transformer? I have a cube that has 40, is it an issue?

[RobsWalker68]

Hi,

The powerplay training guides if I remember rightly recommend a figure around 10, just so the user can see all the dimensions on the screen. 

Personally, I think 40 is on the high side and if I ended up with that many dimensions on the first cut then I would re-appraise my design to make sure I was correctly modelling my process.

Kind Regards


Rob     

[Arsenal]

Robwalker is right - that's too many dimensions. Think of how it would affect performance - Transformer has to roll up the measures at each level of each of those dimensions

I would re-evaluate my design as well

[dssd]

Is it possible for two cubes to be of the same size in terms of GBs and yet take different amounts of build time. Its assumed that there are not abnormalities and hardware being same.

[RobsWalker68]

Hi,

Cube build times will depend on such things as the structure/design of your model, number of categories, the partition strategy as explained by Arsenal in the other post and of course the interaction of the data sources with transformer.

This combination of factors is likely to lead to differing cube build times even though they may have similar sizings.

Rgds

Rob

[cognostechie]

One reason why most people have that many dimensions is because they create dimensions with flat hierarchies (just one level without anything below it) because they think they need those columns to be used in the filters/prompts or to display all values of those items. They think if they put that column as a child in a level they will not be able to see all values but rather see only as many values as applies from the category they are drilling down from. This is a general misconception because even if a column is the 5th level in a dimension, it is till possible to see all values of that level without having to drill down !