Re: Normalizing vs. Denormalizing

A rule of thumb that has worked for me is 'Normalize for storage, un-normalize for reports'. This means normalize your tables (to the 3rd form) for the best performance and most efficient storge. Then create views (un-normalize) that link various tables for the users to access for reporting, data analysis, etc.

This seems simplistic, but it works for me.

Best Regards,

Mike Strong
(713) 274-3317
Texas Instruments, Inc./Houston
Oracle DBA

> epepa_at_aol.com (EPepa) wrote:
> >I'm in the process of helping to develop an Oracle database for a plant
> >that's very concerned with speed. They enter about 5000 orders per month
> >and each order is one record (no detail lines). Since I took the Oracle
> >classes, I have the normalization techniques down pretty well, but my
> >concern is the time that it would take in disk reads in order to access
> >the several tables required to normalize this company's orders. (Each
> >order should store information into several tables that would have 4-5
> >sets of 5-10 pieces of data each, for a total of about 35 denormalized
> >columns.) If I denormalize the data, it would reduce the number of disk
> >reads and all of the information should be fairly contiguous, but it seems
> >to me that that would defeat the purpose of a relational database.
> >
> >Does anyone have any ideas on where you would draw the line?
>
  Received on Thu Mar 21 1996 - 00:00:00 CET