I’m currently researching SQL Server 2008 as a business intelligence solution, and currently looking at Analysis Services (and I’m pretty new to business intelligence as a whole…)
I’m a bit confused by some of the terms in SSAS, particularly the conceptual differences between cubes and MS’s Unified Dimensional Model.
I believe that a cube in SSAS is basically an OLAP cube — dimensions, measures, something that sits between the underlying data source and a business user.
But then that’s kind of what I understand UDM to be as well.
The docs for SQL Server 2005 seem to suggest as much:
“A cube is essentially synonymous with a Unified Dimensional Model (UDM)”.
But then the SQL Server 2008 pages sort of suggest that UDM is a wrapper for both multidimensional data (cubes) and relational data:
“Use the Unified Dimensional Model to provide one consolidated business view for relational and multidimensional data that includes business entities, business logic, calculations, and metrics.”
This blog post suggests similarly:
“UDM provides a single dimensional model for all OLAP analysis and relational reporting needs. So you can use either MDX or SQL”
Is UDM something that sits above cubes? Or are they the same thing? I presume I would develop cubes with the Cube Designer application; what would I develop a UDM with?
A traditional cube is based on a single measure group (fact) with associated dimensiona in a star or snowflake structure.
A UDM (Unified Data Model) allows you to create a multidimensional structure over just about any underlying SQL data model through the use of reference dimensions and many-to-many dimensions.
Note, just because you can doesn’t mean you should slap AS on top of any old structure – it will most likely perform badly. Traditional star schemas are fastest.
It is also sometimes taken to mean the additional SSAS features such as actions and KPIs.