Re: ER Model - different answers for different join paths

Clark Morrow wrote:

> > But Kimball seems to position DM against ER
> > which (I am right) does not make sense at all. Does this make any
 sense?
> >
>
> Kimball positions DM (design) against data warehouse designs done in a
>
> typical OLTP/ER-like design.
>
> If you replace "ER" with "ER style design" and "DM" with "Dimensional
> Design" in Kimball's articles then it is valid to compare the benefits
> of
> "ER style design" with "Dimensional Design."

I read Kimball similarly to xx_at_yy.com, He does seem to position DM against ER. To Quote from his DM course description:

    Data Warehousing in Depth is based on the proven techniques of     dimensional modelling. Unlike E/R modelling, which is intended for     transaction processing performance, dimensional modelling     emphasizes end user understandability, exceptional query     performance and robust extensibility under changing system     requirements.

ER modelling is not "intended for transaction performance" it is simply to get the data right. The fact that data correctly modelled is not really fit for end user consumption is where DM fits in.

People tend to think dimensionally and hierarchically, even though the world is more accurately represented in logical ER diagrams. So when users approach a database, a dimensional model is more understandable.

IMHO a dimensional model that is not based on a good Logical ER model, and a Dimensional Warehouse/Mart that does not derive its data from a good ODS gives users an easily understood pile of garbage.

(Mis?)readings of Kimball's work has led many warehouse developments in this direction

Henry Feinman Received on Sun Aug 17 1997 - 00:00:00 CDT