Re: Discuss Data Design

In article <3fjssn$g02_at_pearl.whoi.edu> mkander_at_whsun1.wh.whoi.edu (Milly Kander) writes:

>Hello Fellow Netters,
 

>Hopefully this will provoke some discussion about data design. I realize
>that any discussion of data structure is dependent on the type of data;
>what I am looking for are rules of thumb that folks have arrived at through
>years of working with Oracle, as well as trial and error.
 

>Given that:
 

>Data is scientific.
>Data remains relatively static.
>Oracle V7 running on a Sun Sparc 10
 

>Queries:
 

>50% of the time users will query for a subset of the data for 1 year.
>50% of the time users will query for a subset of the data for several years.

>So:
 

>Is it more efficient to have one large table containing several years worth
>of a specific type of data,
 

>or
 

>would it be better to break the data up into smaller separate yearly tables?

I think you are better off to store the data in yearly basis, because it is more difficult to maintain a huge table (in terms of on-line backup/recovery).

>Also,
 

>Normalization. What are the current thoughts on de-normalizing data to cut
>down on joins?

From my experience, you should try the *normalise* your database first and see how long the response time is. If your user is satisfied with the system performance, just leave it. If not try to denormalise a bit until you get the acceptable performace. Received on Sat Jan 21 1995 - 16:24:50 CET