Like David and Yasin, I am against the "broad brush" method and use
histograms only in specific cases with demonstrated benefits.
I don't have a testcase, but I have a documented case at a client where
they changed their week(end)ly statistics gathreing from the default
"for all columns size 1" to "for all indexed columns size skewonly" and
had the bottom fall out of a batch job that usually took 1.5 minutes
because one frequently executes sql (1000s of times) went from .01
elapsed to around 80 seconds elapsed.
So be careful whenever you change your statistics gathering strategy.
And that is not limited to histograms.
Charlotte Hammond wrote:
Hi All,
Can I poll the list on their views on the use of histograms? I've previously been quite conservative, only gathering them when we've had a specific case of bad CBO performance due to skewed data. However I was interested in this article mentioning a "broad brush" approach to histogram gathering:
http://www.dba-oracle.com/oracle_tips_all_columns_histograms.htm
Any thoughts to share?
Thank you Charlotte
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com --
//www.freelists.org/webpage/oracle-l
-- Regards
Wolfgang Breitling Centrex Consulting Corporation www.centrexcc.com -- //www.freelists.org/webpage/oracle-l