Re: Which one is good in terms of performance

Thanks Tim, very thorough and detailed work as usual. I am of course 
pleased to see my opinion confirmed, obtained from casual observation 
but not hardened by rigorous testing, that you can get away with a 
rather low sampling rates for table statistics without sacrificing much 
in accuracy, but the same is not true for index statistics and 
histograms and that one should therefore separate the gathering of table 
statistics (including basic column statistics "for all columns size 1") 
from index statistics and histograms. Unless you can afford a full table 
compute.

Tim Gorman wrote:
> Seema and Wolfgang,
> Here is some info I generated recently, to add to the discussion...
> 
> My apologies for the length of the post, but feel free to delete if not
> interested...
> 
> 
-- 
Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com
--
http://www.freelists.org/webpage/oracle-l

Other related posts: