Re: DBMS_STATS compared to the analyze command: weird!

"Niall Litchfield" <niall.litchfield_at_dial.pipex.com> wrote:

>leaving aside the move to degree=4, you have moved from analyze table XXX
>compute statistics; to analyze table compute statistics for table for all
>columns size 10; or in other words from no histograms to one of 10 buckets
>on every column. it isn't entirely surprising that the performance is
>different.

You didn't get my message right. The performance problem appeared when using a more simple form of the dbms_stats command, without histograms.

>changes in 9.

Funny that Oracle discourages the use of the analyze command and still keeps it internally.
>well you have said method_opt=> for all columns... all that means is collect
>histograms on all columns with the size I specify - hence the equivalence to
>the analyse I had above. Your original statement didn't collect histograms,
>the dbms_stats one did - for every column.

I did not ask to explain my own command to me ;-)

My question is about the impact of using different methods for gathering statistics, which Oracle itself shuffles at its will.

Bye
Rick Denoire Received on Thu Feb 26 2004 - 17:17:05 CST