Re: Normalization question
Date: Fri, 28 Jun 2002 07:55:14 -0400
Message-ID: <3D1C4EA2.A9925F82_at_erols.com>
Steve Kass wrote:
>
>,,,
> Another example might be a database of Australians with
> "preferred language" as an attribute. It may well be that 95%
> of the entries list English for this attribute. If so, one could
> save space by eliminating the attribute and maintaining a
> separate table with one row for each non-English speaker
> containing his or her preferred language. Similarly, one could
> do this and save space for any attribute with a very prevalent
> mode.
>
Steve:
Given the cost of disk storage space I think you would be better off optimizing for performance and ignore disk space consumption. As a DBA I have never heard users or management complain about needing more disks, but I have frequently heard complaints about performance. Performance complaints are best resolved by minimizing the number of disk accesses required to handle a transaction. In the example you cite every transaction would require reading the separate table which, in 95% of the cases, is a waste of time.
Moving from being practical back to theory; the existence of the additional table will probably result in redundant data being carried in the system. As a DBA I have spent thousands of hours locating and ruthlessly eliminating redundant data. This is because all too often I have seen garbage being produced because programs were changed but the data was not corrected in every file in which it appeared.
Jerry Received on Fri Jun 28 2002 - 13:55:14 CEST