You might want to read up on clustered tables and see if this will work for
you ...
This way you'll have logically multiple tables that are physically stored in
the same datablocks...
"Steffen Ramlow" <s.ramlow_at_gmx.de> wrote in message
news:a8hm7r$sqgtj$1_at_ID-54600.news.dfncis.de...
> what r the advantages / disadvantages of this?
>
> sample:
>
> Main (pk, id, c1, c2)
> Sub1 (pk1, fk1, c10, c11)
> Sub2 (pk2, fk2, c20, c21)
>
> rows:
>
> Main: 1,1,2,3
> Sub1: 1,1,4,5
> Sub2: 1,1,6,7
>
> vs.
>
> BigMain(pk,id,c1,c2,c10,c11,c20,c21)
>
> rows:
>
> 1,1,2,3,4,5,null,null
> 2,1,2,3,null,null,6,7
>
>
> the rows are always read as when Main and Subx would be inner joined
>
> obvious is, that BigMain has many null values (there are up to 10 sub
> tables) but u do not need to join Main and Subx
>
> i would use Main + Sub, coz it is better to maintain and to tune, but what
> about the costs of the join?
>
> both tables may contain millions of rows
>
> what r ur options?
>
>
>
Received on Thu Apr 04 2002 - 15:09:05 CST
