Re: does anybody recomends using dbms_stats.gather_schema_stats with for columnns size skewonly

  • From: Wolfgang Breitling <breitliw@xxxxxxxxxxxxx>
  • To: aleon68@xxxxxxxxx
  • Date: Sun, 16 Jan 2005 17:08:26 -0700

comments inline

At 03:15 PM 1/16/2005, you wrote:
>We are estimating statistics for each schema of our aplication using
>with method_opt FOR COLUMNS SIZE SKEWONLY (not using the ALL) and
>estimating with dbms_stats.auto_sample_size, options GATHER and CASCADE 
>does this approach will estimate the estadistics for all columns using
>size 1 if the data is not skewed? would it do it only if needed?

This approach - method_opt=>'for columns size skewonly' - will perform very 
well since all it is doing is counting the number of rows of the table and 
update the num_rows and blocks statistics in dba_tables. OK, if you use 
cascade=>true it will also gather index statistics. What it does NOT do is 
gather any column statistics nor even the avg_row_len statistic for the table.

My recommendation is not to waste any computing resources on that approach.

>Alfonso Leon
>Guayaquil - Ecuador


Wolfgang Breitling
Centrex Consulting Corporation 


Other related posts: