Re: Normalization by Composing, not just Decomposing
Date: Tue, 13 Apr 2004 15:45:05 -0500
This terminology is just peachy -- I'll try to stick to it better. Dimensional modeling --> Star Schema design (whether or not it is for a relational database) --> Cube specification --> Cube construction.
It seems to me that the dimensional modeling is more like relational modeling than ER modeling (which is more conceptual) but that is not a major point and if I'm wrong on that small point, it pains me naught. Thanks. --dawn
"Laconic2" <laconic2_at_comcast.net> wrote in message
> The terminology, by Kimball, refers to dimensional modeling. I would put
> dimensional modeling on a par with ER modeling. It's abstract enough so
> that it's not particularly tuned to any implementation.
> A star schema is a dimensional model mapped into tables. This makes it
> to imlpement in something like Oracle or DB2.
> A cube is a dimensional model mapped into some kind of multidimensional
> database system. Cognos powerplay is one example.
> You can realize a dimensional model by building a cube, just as you can
> realize an ER model by building a database.
> However, it may be better to do the modeling first, then the building,
> rather than the other way around, especially for large projects. But,
> first, we tend to do things backward. We learn how to program before we
> learn how to design programs, and we learn ho to design programs before
> learn how to analyze requirements. Funny the way we think, huh?
Received on Tue Apr 13 2004 - 22:45:05 CEST