Re: Partitioning or BitMap-Index?

On Wed, 12 Feb 2003 13:00:16 -0000, "David Kurtz" <info_at_go-faster.co.uk> wrote:

>If you are updating this column then partitioning is probably not a great
>idea because you would move the row from one partition to the other. Lots
>of extra work involved.
>Similarly, if you are regularly undating the table a bitmap index may not be
>a good idea either. Updating the index can be expensive and the index can
>grow very rapidly.
>
>Are you using this column as a flag to find the rows that you want to
>process? In which case have you considered what is sometimes called a
>'sparse index'.
>If the column had values 1 and NULL (and most of the rows had value NULL in
>this column), then only the rows with value 1 would appear in the index, and
>the index would efficiently identify the rows.
>
>If you have to use not null values in all rows then every row will be in the
>index. A histogram may also help if the data is mostly in one value, but
>only if the queries use a litteral value on this column.

First of all, thanks a lot for your help.

I am currently using this column with possible values of 0 and 1. The data in this column is onyl created when the dataset is created in the table. There are no DML-Statements on this column and the column is used in every SQL-Statement witch run against the table in the where-clause.

What do you think about it?

Regards,

Christian Received on Thu Feb 13 2003 - 00:25:28 CST