Check out the book Enterprise Data Modeling, available from Amazon at:
Physical modeling, like indexing and denormalization (which is justified in
1/10 of the instances it occurrs), depend entirely upon your application and
usage of the data. Most of the sql server dba books will have at least some
information on indices and their impact. The general rule of thumb is that
adding indices makes data retrieval faster, but trades off with insert time
and storage space. Read any book by C.J. Date.
The best education, in my opinion, comes from experience. This doesn't
necessarily mean doing it as your job, but that helps. Think of an
application, for example a recipe database. Then try to normalize it.
Think about the dependencies, think about the impact of table structures.
Think about correlations and where an attribute really resides.
> Does anyone know if there are resources with best practices for
> physical data modelling
> I'm thinking of
> - how many indexes
> - valid reasons to denormalize ?
> - how many columns in a table ?
> - what to think about when defining foreign key constraints ?
> - how to implement roles (implement doctors and patients as roles of
> "people" because they both have name and address fields ?
> Things like that.
> I do not expect specific answers from you, but a way to find them
> Thanks a lot
> Frieda van der Sar