Re: SQL Profiles in 10g

I haven't worked this through in detail, but I think that the strategy used by dbms_sqltune is something like:

    start with the 'normal' execution plan     spend extra time looking for a better plan     check for better stats
    check the cost of the normal plan with better stats     look for better plans with the better stats     repeat for hypothetical situations such as:

        if X were true, then A would be a legal transformation
        of the query (e.g. extra columns declared not null)
        check cost of new transformation with old and new
        statistics.

Possibly you've found a plan where

    checking more options doesn't help
    changing the stats doesn't change the choice of plan     new, hypothetical, transformations don't help.

It seems a little unlikely - but I guess it could happen. "Nothing I can do will improve the situation" is sometimes going to be the 'mechanically' correct answer.

-- 
Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk/faq/ind_faq.html
The Co-operative Oracle Users' FAQ

http://www.jlcomp.demon.co.uk/seminar.html
Optimising Oracle Seminar - schedule updated Sept 2nd





"Christian Antognini" <christian.antognini_at_trivadis.com> wrote in message
news:41378a31$1_at_post.usenet.com...


> **** Post for FREE via your newsreader at post.usenet.com ****

>

>

> "Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message

> news:ch7vg5$qv$1_at_hercules.btinternet.com...

> >

> > Did you get anything in the "Recommendations" or

> > "Restructure" section ?

>

> Hi Jonathan

>

> No recommendations at all, i.e. only the message "There are no

> recommendations to improve the statement.".

>

> Chris