Re: Weird OPTIMIZER behavior for SQL*Net queries

This looks as if it should be impossible.

The only idea I can think of at the moment is that there is a glogin.sql, or login.sql script that gets executed in one set of circumstances and not in the other. (which shouldn't happen anyway, but the only way to get two different paths from the same server is to have two different sets of optimisation parameters, and the only way to change them dynamically is through ALTER SESSION).

A second thought - perhaps there is an nls_lang (o/s) parameter set that makes one index undesirable on one connection but not the other - I don't suppose the tnsname entry has a funny ENVS= entry ?

--

Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk

Jerzy Tomasik wrote in message <37D0A5B2.BC9E5FE6_at_tus.ssi1.com>...
>Hi,
>
>I'm seeing a strange performance problem. The same query performs very
>well when executed
>with a direct connection (e.g. sqlplus scott/tiger) and very poorly when
>executed through SQL*Net
>(e.g. sqlplus scott/tiger_at_mydb). Running the analyzer shows that the
>execution path is different in
>both cases.
>BTW, the query generates virtually no network traffic (returns just a
>count).
>
>Is there any way I can control or force the optimizer to behave the same
>way for a query
>coming through SQL*Net as it does for local queries?
>I've read the docs about the hints, etc, but I haven't seen anything
>that seems to be the answer.
>
>TIA,
>
>-Jerzy
>
Received on Mon Sep 06 1999 - 06:41:31 CDT