| Well,
The presentation mentions the problem and the way it should be solved.
The way is to look for common dimensions, or generic dimensions like
they are called in the Solution Architecting Data Warehouses workshop.
Ralph Kimball uses the same terms in his excellent class on data
modeling for the Data Warehouse. He even says it a bit stronger: the
Data Warehouse is a conglomerate or cluster of individual Data Marts,
tied together by common dimensions.
Question remains: how do you develop these generic dimensions in order
to have scaling. The answer is that you need some architectural
blueprint to achieve that, where common dimensions are discovered when
developing the technical view (subject area's overview model). This
model is (and should be) closely linked to the value chain of the
organization. In the workshop mentioned above this process is tought.
The problem as described in the presentation is very real. I just
returned from a visit at a large company in Israel where this problem
manifested itself. The customer saw no escape besides:
- delaying, possible stalling the current implementation of a DM in one
of the departments, caused others jumping on the wagon
- developing seperate DM's, with great chance of leading to islands of
data
Digital offered help by doing a short (say 3 days) architecture
workshop in parallel to re-synchronise and discover generic dimensions.
We are in the unique position to be able to do this.
Cheers,
Henny
|
|
Henny,
You are exactly right and in the Rams training for Data marts we
have just extended with a design part we are trying to address this step.
What is built in front of the Logical (star-schema in this case) is
a conceptual design where the business view (also called subject
data-model) is described. The intent of this high-level modelling is to
catch the common dimensions, model the output view and check it against the
data sources to make the project safe (especially for a 6 weeks
project : early in the process). Ralph Kimball told me that he usually
spends 5-6 weeks on it for a full Data-warehouse at the beginning of
the process: for the Ddm we are only focusing on one subject-area.
The next DataMart will be consolidated with the first one built (to
share the common dimensions) and even a migration to a Datawarehouse would
re-use this conceptual design.
Cheers
Pierre
|