If I have a key comprised of Column-1 plus Column-2, and if I
Select X where Column-2 = "yyy"
then I suppose that the retrieval takes place on a pure
sequential basis just as if I were selecting on some column
not part of any key. Is that right?


Ross
Vontay, Virginia
http://community.webshots.com/user/ross_klatte
http://www.geocities.com/foundlingfather/

>If I have a key comprised of Column-1 plus Column-2, and if I[color=blue]
>Select X where Column-2 = "yyy"
>then I suppose that the retrieval takes place on a pure
>sequential basis just as if I were selecting on some column
>not part of any key. Is that right?
>
>[/color]

p.s.
What is the objection, if any, of adding another index comprised
of Column-2 plus Column-1?


Ross
Vontay, Virginia
http://community.webshots.com/user/ross_klatte
http://www.geocities.com/foundlingfather/

"Ross Klatte" <klatteross@aol.commmm> wrote in message
news:20040516145559.04109.00000673@mb-m21.aol.com...[color=blue]
> If I have a key comprised of Column-1 plus Column-2, and if I
> Select X where Column-2 = "yyy"
> then I suppose that the retrieval takes place on a pure
> sequential basis just as if I were selecting on some column
> not part of any key. Is that right?
>
>
> Ross
> Vontay, Virginia[/color]

In the case you described above, DB2 will "usually" do a index scan rather
than a tablespace scan of the data pages. So that means DB2 will read all
the index leaf pages sequentially instead of using the b-tree of the index.

Instead of creating another index on (column-2, column-1) just create a new
index on column-2. Column-1 in this new index would be redundant if you keep
the first index (column-1, column-2).