Mailing-List: contact ojb-dev-help@db.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "OJB Developers List" <ojb-dev@db.apache.org>
Content-Type: text/plain;
  charset="iso-8859-15"
From: Oleg Nitz <on@ukr.net>
Reply-To: on@ukr.net
To: "OJB Developers List" <ojb-dev@db.apache.org>
Subject: Re: Performance decrease of ODMG-api in CVS
Date: Sun, 2 Nov 2003 18:31:49 +0000
User-Agent: KMail/1.4.2
References: <oprwhnt6yl2o6nlo@smtprelay.t-online.de>
In-Reply-To: <oprwhnt6yl2o6nlo@smtprelay.t-online.de>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Message-Id: <200311021831.49569.on@ukr.net>

Hi Armin,

I found the reason of the "performance decrease": I changed the ODMG test=
 on=20
2003/09/13, I added two lines marked by plus below:

        public void updateArticles(PerfArticle[] arr) throws Exception
        {
            m_tx.begin();
            for (int i =3D 0; i < arr.length; i++)
            {
                m_tx.lock(arr[i], Transaction.WRITE);
+               arr[i].setArticleName("" + System.currentTimeMillis());
            }
            m_tx.commit();
        }

        public void updateArticlesStress(PerfArticle[] arr) throws Except=
ion
        {
            for (int i =3D 0; i < arr.length; i++)
            {
                Transaction tx =3D odmg.newTransaction();
                tx.begin();
                tx.lock(arr[i], Transaction.WRITE);
+               arr[i].setArticleName("" + System.currentTimeMillis());
                tx.commit();
            }
        }

Previously the update test didn't perform database UPDATE because objects=
=20
remained unchanged, so comparison with other APIs where incorrect.
After I added these lines to the OJBPerfTest.java in the CVS state as on
August 1, the ODMG test took 30% longer then before.

Regards,
 Oleg

On Friday 03 October 2003 22:38, Armin Waibel wrote:
> Hi all,
>
> some user reported about a ODMG-api performance
> decrease of current CVS. I think they are right.
> The performance test (perf-test task) for odmg-api
> took 43 sec with rc4 (27 July)
> took 37 sec with new DList implementation (1 August)
> took 55 sec (ca. 25 Sep)
> took 48 sec after fixing lock cleanup (today)
>
> So performance of odmg-api decrease more than
> 30% between 08/01 and 09/25!
> I couldn't find out the reason for that. Seems that
> the changes made in odmg classes between this
> period are minor. Maybe the reason are changes in
> the kernel. But the kernel only performs ca. 5%
> slower.
> Any comments, proposals, solutions?
>
> regards,
> Armin
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: ojb-dev-unsubscribe@db.apache.org
> For additional commands, e-mail: ojb-dev-help@db.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: ojb-dev-unsubscribe@db.apache.org
For additional commands, e-mail: ojb-dev-help@db.apache.org