Re: Bind variable peeking and Dynamic sampling

Fairlie

I see a few potential problems with using dynamic sampling in your scenario:
a) unless you can use hints in the sql you'll need to delete the statistics on the table for dynamic sampling to kick in 
b) do you really want to do sampling on 400 txn/sec?
c) the statement may not get reparsed as long as the plan is in SGA and valid (never used DS in a scenario like that so don't know for sure) which would alleviate the concern under b), but wouldn't change the situation vs. peeking d) AFAIK, dynamic sampling in 10gR1 isn't quite up to snuff for index and column sampling. It uses the same defaults the CBO has been using for tables without statistics since its inception which means you get the same column selectivity regardless of the predicate value. Only 10.2(.0.3) finally fixes that.
You'll need at least ds=2, which is already the default. If I'd go with something like that I'd probably use ds=4 

As always, there is no substitute for testing.

At 02:43 AM 5/8/2007, fairlie rego wrote:

Hi,
I am at a site which is a high intensive OLTP site doing more than 400 txns/sec on a 8 node cluster on 10.1.0.5 RAC. We are thinking of switching off bind variable peeking and looking into an optimal value of dynamic sampling since we have plans flipping all the time and causing havoc.
I would like to know if there are any other customers who have disabled bind variable peeking and what value of dynamic sampling is being used by them 

Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com 


______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email ______________________________________________________________________ 
--
http://www.freelists.org/webpage/oracle-l

Other related posts: