Re: Performance Problem but no Performance Tuning Tool
- From: Martin Berger <martin.a.berger@xxxxxxxxx>
- To: ocat31@xxxxxxxxxx
- Date: Wed, 25 Nov 2009 10:09:22 +0100
Sam,
unfortunately I cannot give you any tool, but just some questions to start
(maybe you have answered these for yourself, but I cannot find them here)
* do you have any kind of baseline, numbers, metrics, traces, whatever from
the time before you changed the system? (to compare your current numbers
against)
* do you have a valid measurement of your busines processes/SQLs and
* do you have a wel ldefined goal how fast they should run?
* Can you identify the worst process/SQL in your system, so you can start
there?
with all these preparations (I hope I have not left out something, but I'm
sure in this case others will help) a structured investigation might be
possible.
Dependent of some of the answers, I'd enable sql-trace on that particular
session/module/whatever, or check some session or in worst case system
statistics.
Even without ASH, (semi-)manual sampling on v$session (waits, events) might
also help.
Once again I can recommend tanel poders snapper for such collections.
sorry, no short answer, but maybe a possible path to follow.
Martin
On Wed, Nov 25, 2009 at 05:06, Sam <ocat31@xxxxxxxxxx> wrote:
>
> Hi all,
>
>
>
> I work for an organization that has no performance tuning tools, and one of
> my jobs is to tune our enterprise Oracle 10.2.0.4 database on AIX 5.3. A
> few changes were made to our environment on a long weekend in October, and
> since that time response time has been a lot worse. We had a consultant
> come in for a couple of weeks, but performance problems remain. Here are
> the changes made on that weekend:
>
>
>
> 1) Move the databases to more powerful hardware. We remained at AIX 5.3.
>
> 2) Upgrade from 10.2.0.3 to 10.2.0.4
>
> 3) All unnecessary network ports were closed (it took a while to get all
> the needed ports open)
>
> 4) Prior to the upgrade, our primary enterprise database had its own set of
> dedicated SAN disks. I don't know how many spindles. Other databases were
> on different SAN disks. After the upgrade, the SAN consisted of 16
> spindles, and all our Oracle databases are spread across all 16 spindles;
> i.e., there are no longer dedicated disks just for the enterprise database.
> I haven't been told the RAID type has changed, so for now I assume it is the
> same.
>
>
>
> If anybody has scripts that can help isolate the root cause of the general
> performance degradation that would be most helpful. I am familiar with
> Oracle wait event analysis, but without access to a Performance Tuning tool
> I do not have a good method of pin-pointing what the problem waits are or
> what is causing the waits to exist.
>
>
>
> I know dbConsole comes with a very good tuning pack, but our free trial
> period is over. Also, I think that using the ASH scripts requires an
> additional license, so that is also not an option for me.
>
>
>
> So if somebody has scripts to share, that would be helpful to me.
>
>
>
> Thanks very much,
>
>
>
> SB
>
>
--
Martin Berger martin.a.berger@xxxxxxxxx
Lederergasse 27/2/14 +43 660 660 83306
1080 Wien http://berx.at/
Sent from Wien, Österreich
Other related posts:
