Hi all
With respect to database design and performance is it good to have more
tables with fewer records or with few tables having lots of records.
Anil
Nick Landsberg - 30 Sep 2004 18:19 GMT
> Hi all
>
> With respect to database design and performance is it good to have more
> tables with fewer records or with few tables having lots of records.
>
> Anil
What are your access patterns?
Do most of the queries access a small percentage
of the attributes or do they access a large
percentage of the attributes?
There is no general answer, it all depends
on usage patterns. Remember, 80% of life
follows the 80/20 rule. Determine the 20%
of your queries that are run 80% of the time,
and base your decision on this.
NPL

Signature
"It is impossible to make anything foolproof
because fools are so ingenious"
- A. Bloch
Gene Wirchenko - 30 Sep 2004 20:27 GMT
>With respect to database design and performance is it good to have more
>tables with fewer records or with few tables having lots of records.
It is better to have normalised tables. Whether this means more
or fewer is largely irrelevant.
Sincerely,
Gene Wirchenko
Computerese Irregular Verb Conjugation:
I have preferences.
You have biases.
He/She has prejudices.
Laconic2 - 30 Sep 2004 20:30 GMT
> Hi all
>
> With respect to database design and performance is it good to have more
> tables with fewer records or with few tables having lots of records.
It's good for the logical design to be logical. Put things in the same
table if they belong together, and let the chips fall where they may. Those
who sacrifice logic for the sake of good performance end up with neither
logic nor good performance. (apologies to Ben Franklin).
--CELKO-- - 01 Oct 2004 00:29 GMT
>> With respect to database design and performance is it good to have
more tables with fewer records [sic] or with few tables having lots of
records [sic]. <<
The question makes no sense. Rows are not records; many modern VLDB
products (see Nucleus from Sand for example) compress the data if it
is repeatative, so what is materialized at run time takes up very
little physical storage but can be used to build millions of rows.
You are still thinking that a RDBMS has to work like a traditional
file system.
Worry about getting the design right first; performance is not as
important.