Re: Stats gather job for PeopleSoft Tools version 8.4.8

  • From: "John Darrah" <darrah.john@xxxxxxxxx>
  • To: "Wolfgang Breitling" <breitliw@xxxxxxxxxxxxx>
  • Date: Wed, 25 Apr 2007 14:44:09 -0600

Hey guys, I didn't get a chance to thank you both for your suggestions.
Currently we are looking at leaving the tables unlocked and writing a custom
script to gather stale on the non transient tables.  The reason for this is
that it seems that gathering stats still produces superior information to
dynamic sampling when it comes to index stats.  I only have anecdotal
evidence to support this theory, I haven't run 10053 traces with stats and
without but I have seen better plans produced with stats gathered than
without and the missing piece seemed to be clustering factor.  Take all of
this with a grain of salt as I have not tested but just thought I'd pass
this along.  Have either of you ever tried running peoplesoft with cursor
sharing set to exact or similar?  Just wondering.

Thanks again,


On 4/17/07, Wolfgang Breitling <breitliw@xxxxxxxxxxxxx> wrote:

At 09:49 AM 4/17/2007, David Kurtz wrote:

>estimate_percent=>1, method_opt=> 'FOR ALL COLUMNS SIZE
>estimate_percent=> dbms_stats.auto_sample_size, method_opt=> 'FOR ALL
>I don't approve of FOR ALL INDEXED COLUMNS SIZE 1 - that will destroy any
>histograms that have been defined.

I disapprove of that as well, but for a different reason. I do not
approve of the "for all INDEXED columns" restriction. As if only
indexed columns needed statistics. The "size 1" part destroying
histograms is no different from the analyze behaviour of prior releases.


Wolfgang Breitling
Centrex Consulting Corporation

Other related posts: