Re: Odd join selectivity RESOLVED (partially)

From: Greg Rahn <greg_at_structureddata.org>
Date: Fri, 7 Mar 2008 12:37:37 -0800
Message-ID: <a9c093440803071237keb9cd70k400f204b93b5997a@mail.gmail.com>


This white paper may be of use:
Designing Applications for Oracle Label Security http://www.oracle.com/technology/deploy/performance/pdf/design_for_ols.pdf

On 3/7/08, Charles Schultz <sacrophyte_at_gmail.com> wrote:
> After looking at all the information collected, Oracle Support has responded
> by saying that the optimizer is actually taking a wild shot in the dark with
> this selectivity number; the root of the issue is that we have VPD enabled
> on the underlying tables, and since SYS_CONTEXT is non-deterministic, the
> run time optimizer is not correctly determining the selectivity; more
> importantly, the run time optimizer is unable to calculate the proper
> number. I still have no idea how the number was actually generated; for all
> intents and purposes, it could come from a random number generator, I guess.
> =) The suggested work-around is to have SQL Tuner come up with a profile. I
> know one can assign statistics to functions using the Data Cartridges, but
> that seems like overkill, and the last time I plunged into that murky bog, I
> came out with no noticeable benefits.
>
> Thanks for all those who responded.
>

-- 
Regards,

Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l
Received on Fri Mar 07 2008 - 14:37:37 CST

Original text of this message