Re: Default optimizer_index_caching value DUMB ...

Jonathan Lewis wrote:

> A point that is often overlooked with optimizer_index_caching
> is that it supposed to apply only for index costs during nested
> loops (although it does seem to be applied for inlist iterators).
>
> Consequently it is not really a measure of what fraction of
> an index is cached, it is a measure of how many rows I
> select from a driving table that require me to collect the
> same data from the inner table repeatedly.
>
> For example:
> If I select 5 rows from table A and then
> join to table B, it is possible that I will do a
> physical read for the index blocks for tableB
> after reading the first row from table A - but
> then use the same index blocks for the join to
> the next 4 rows from table A.
>
> In this case, an optimizer_index_caching value
> of 80% would be the representative value.

Thanks.

-- 
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace 'x' with 'u' to respond)