Re: Performance different between two users

Sybrand Bakker <gooiditweg_at_nospam.demon.nl> wrote in message news:<hk2e2vsh8snmpi3i9u791rcelppikhjun5_at_4ax.com>...
> On Thu, 16 Jan 2003 19:21:38 -0000, postmaster_at_cjc.org (Cheng-Jih
> Chen) wrote:
>
> >
> >Hi, we're developing an application, and have several user logins set
> >up for Oracle for testing.
> >
> >Each of these users are created identically, and have the share the same
> >default tablespace and temporary tablespace. They all have the same
> >rights, and their schemas are created from the same script. The data
> >in these schemas are identical.
> >
> >However, we're seeing performance differences between two of the users,
> >where one user is consistently two or three times faster than the
> >other user.
> >
> >Any suggestions on where to look? I've checked the data segments that
> >they own, and these seem to be identical in terms of fragmentation, etc.
> >Indexes, and so on, look to be the same, since the schemas were generated
> >automatically. I'm not sure where else to look.

Are they coming in as sqlnet clients, or are they on the same machine?

Maybe the data segments aren't as identical as you think - perhaps a high water mark got set in one? Try truncating and reloading the data. Even check the table row counts.

Be sure the explain plans are the same. Re-ANALYSE all tables (or be sure you don't have any analysis if you expect RULE based and have set CHOOSE). Any dblinks involved? Sometime rights and synonyms can look the same, but...

Also, it's possible (depending on your hardware) that identical looking segments are in places with differing response times - for example, a bad spot on disk may eventually work after several retries.

Only one instance? platform and versions? How big are buffers?

> >
> >Thanks for your help.
>
> Your users will both parse the same statements, as different schemas
> are considered, and likely they will both push each others data out of
> the buffer cache.

If that were the problem, wouldn't the effect teeter-totter between the users?

>
>
> Sybrand Bakker, Senior Oracle DBA
>
> To reply remove -verwijderdit from my e-mail address

jg

--
@home is bogus.
Little is actually known about people with hyperthymia, for the simple
reason that they don't see psychiatrists complaining that they are
happy.