Re: Bitmap Index

A good rule of thumb is if the number of distinct values you would be indexing make up less than 2-3 percent of the total number of rows. For example, if you have a table with 100,000 rows, you could have a bitmap index of 2-3 thousand values. We used it for 4200 distinct values in a table of 30 million rows. One other thing, in 7.3 we found that to make the optimizer work correctly, every time we went to analyze the table, we first had to delete statistics, and then estimate statistics. If we did not delete first, the optimizer did not work as it should. I hope that this helps.

Alan V.

Marcel Romahn wrote in message <34EAAB1C.28E5B8C9_at_messer.de>...
>Conditions for using bitmap indexes essentialy are the low cardinality
>of the indexed columns with low selectivity, which means huge amout of
>retrieved rows.
>
>What is "low cardinality" in this context. Yes, a GENDER or YES/NO
>column is easy to understand. But what about attributes with 20 or 30
>different values. Where is the break even point in comparison with
>b-tree indexes.
>
>Thanks in advance
>
>Marcel Romahn
>
Received on Wed Feb 18 1998 - 00:00:00 CST