Cubes or not - Urgent

[bloggerman]

How do you guys decide if we should build cubes or not looking at  requirements

[wyconian]

Hi

That's a pretty big question :-)

In a nutshell I would use a cube if the requirement calls for aggregated data and wants to look at some kind of period on period analysis or wants to comapre measures e.g. actual sales against budegt sales etc

Good luck

[bloggerman]

thanks...Its a big dilemma too

[tjohnson3050]

Also performance is key.  We used to start with dmr models in framework manager and if performance became an issue, then we would build a cube. 

Dashboards usually perform much better with cubes.

[wyconian]

Performance on a cube is pretty much always better than reporting straight against a warehouse because cubes tend to be more focused and a smaller data set than reporting against the whole warehouse.   Also cubes, by definition, have aggregated measures rather than having to roll measures up.

Personally I would avoid DMR.  Creating heirarchies, dimensions, aggregated measures etc on the fly obviously has inherant performance issues.  If you want to use DMX functions build a cube. :-)