BTW If the first scenario is important for you please note that most of the
hotspots in EPP 5.2.2 are due to NavigationImpl.getIndex() maybe the way it
is implemented in Gatein should be reviewed?
Please find JProfiler hotspots as attached file
On Fri, Nov 2, 2012 at 4:22 PM, Nicolas Filotto <
nicolas.filotto(a)exoplatform.com> wrote:
I launched your first scenario on EPP 5.2.2 GA now I'm stuck
since I don't
have JPP to be able to compare.
On Fri, Nov 2, 2012 at 10:38 AM, Michal Vanco <mvanco(a)redhat.com> wrote:
> Forget to attach files...
>
> ----- Original Message -----
> > From: "Michal Vanco" <mvanco(a)redhat.com>
> > To: "Nicolas Filotto" <nicolas.filotto(a)exoplatform.com>
> > Cc: "Nick Scavelli" <nscavell(a)redhat.com>,
"gatein-dev" <
> gatein-dev(a)lists.jboss.org>, "Honza Fnukal"
> > <hfnukal(a)redhat.com>
> > Sent: Friday, November 2, 2012 11:37:04 AM
> > Subject: Re: [gatein-dev] Performance degradation between JCR 1.14 and
> 1.15?
> >
> > Hi,
> >
> > ----- Original Message -----
> > > From: "Nicolas Filotto" <nicolas.filotto(a)exoplatform.com>
> > > To: "Michal Vanco" <mvanco(a)redhat.com>
> > > Cc: "Nick Scavelli" <nscavell(a)redhat.com>,
"gatein-dev"
> > > <gatein-dev(a)lists.jboss.org>
> > > Sent: Friday, November 2, 2012 10:23:26 AM
> > > Subject: Re: [gatein-dev] Performance degradation between JCR 1.14
> > > and 1.15?
> > >
> > > Good, next questions:
> > >
> > >
> > > 1. Where can I get JPP 6.0.0 ER02? Is this
> > >
>
http://intranet.exoplatform.org/rest/jcr/repository/collaboration/Groups/...
> > > what you meant?
> >
> > No, this is plain EAP, I'm talking about JPP 6.0.0 ER02 build on top
> > of EAP 6.0.1 ER03 - Honza should be able to share on intranet with
> > you.
> >
> > > 2. What are the steps to reproduce?
> >
> > To reproduce first scenario:
> > - use attached files
> > - add to configuration.xml
> > <!-- Configuration for many pages and navigations -->
> > <import>war:/conf/portal/manyPagesPortal-configuration.xml</import>
> > - place manyPagesPortal-configuration.xml at conf/portal folder
> > (using NewPortalConfigListener)
> > - copy folders with generated pages/navigations to portal folder
> > - copy nullPortletPortalPage.war at standalone/deployments
> > (contains very simple hello world portlet which is referenced at
> > pages.xml)
> > - (I have some extra pages for groups/users, but it's not necessary
> > to add)
> > - start the portal
> > - then do the same with epp 5.2.2 to measure difference
> > - difference is at initialization time, then you can verify
> > initialized portals - there is big improvement in times when
> > accessing portals thanks to new page service I suppose
> >
> > To reproduce second scenario:
> > - use the write portlet I have attached last time - place them
> > somewhere at portal and run load test on this page
> > - jcr operation (with randon nodes write) is called everytime
> > portlet is being rendered, you should be able to compare difference
> > with current/old/updated version of portal/jcr
> >
> > Regards,
> > Michal
> >
> > >
> > >
> > > On Fri, Nov 2, 2012 at 9:11 AM, Michal Vanco < mvanco(a)redhat.com >
> > > wrote:
> > >
> > >
> > > Hi Nick,
> > >
> > > I'm using MySql 5.5 for portal performance.
> > >
> > >
> > > Best regards,
> > > Michal
> > >
> > > ----- Original Message -----
> > >
> > >
> > > > From: "Nicolas Filotto" < nicolas.filotto(a)exoplatform.com
>
> > > > To: "Michal Vanco" < mvanco(a)redhat.com >
> > > > Cc: "Nick Scavelli" < nscavell(a)redhat.com >,
"gatein-dev" <
> > > > gatein-dev(a)lists.jboss.org >
> > > > Sent: Friday, November 2, 2012 10:07:12 AM
> > > > Subject: Re: [gatein-dev] Performance degradation between JCR
> > > > 1.14
> > > > and 1.15?
> > > >
> > > > Michal, same question for you, which db did you use?
> > > >
> > > >
> > > > On Thu, Nov 1, 2012 at 12:09 PM, Michal Vanco < mvanco(a)redhat.com
> > > > >
> > > > wrote:
> > > >
> > > >
> > > > Hi,
> > > >
> > > > I have some feedback on this topic as well.
> > > >
> > > > I have tried 2 scenarios:
> > > > 1) initialization of portal with many pages (server startup time)
> > > > 2) jcr portlet (for writing some randon data in jcr) - portlet is
> > > > accessed with concurrent users during perf test
> > > >
> > > > ad 1)
> > > > 3 portals with many pages at different structure are initialized
> > > > (each portal has +- 780 nodes/pages - [size,height]: [5,4] ,
> > > > [27,2]
> > > > , [780,1])
> > > > - EPP 5.2.2 started at 50m:51s:425ms
> > > > - JPP 6.0.0 ER02 started at 5154494ms = 85,9m
> > > > - JPP 6.0.0 ER02 with upgraded jcr core started at 5054052ms =
> > > > 84,2m
> > > >
> > > > Accessing these portals is much better at JPP 6 as I have already
> > > > mentioned at previous email.
> > > >
> > > > ad 2)
> > > > This is a refreshed scenario we have tested, textlogs from perf
> > > > runs
> > > > are attached - best results were with epp 5.2.2
> > > > For jpp 6, results were slightly better with upgraded jcr core
> > > > I'm also attaching jcr portlets which can be reviewed - write
> > > > portlet
> > > > is used, it adds some random data at jcr when portlet is accessed
> > > >
> > > > Please keep in mind that it's a comparison of 2 major versions
> > > > which
> > > > are quite different, but I would expect these scenarios do not
> > > > differ too much.
> > > >
> > > > Best regards,
> > > >
> > > > Michal
> > > >
> > > > ----- Original Message -----
> > > >
> > > > > From: "Nick Scavelli" < nscavell(a)redhat.com >
> > > > > To: "Nicolas Filotto" <
nicolas.filotto(a)exoplatform.com >
> > > > > Cc: "Michal Vanco" < mvanco(a)redhat.com >,
"gatein-dev" <
> > > > > gatein-dev(a)lists.jboss.org >
> > > > > Sent: Tuesday, October 30, 2012 11:10:03 PM
> > > > > Subject: Re: [gatein-dev] Performance degradation between JCR
> > > > > 1.14
> > > > > and 1.15?
> > > > >
> > > >
> > > >
> > > > > Ran some more tests as I saw 1.15 having better numbers for
> > > > > pages
> > > > > less then 500.
> > > > >
> > > > > Looks like the new jar is scaling better, and even
> > > > > outperforming
> > > > > 1.14
> > > > > except for 500 pages. Here are the results
http://bit.ly/T2ghPK
> > > > > .
> > > > > The second tab shows the results for batching commits rather
> > > > > then
> > > > > waiting and committing 500 pages at once. 1.15 outperforms 1.14
> > > > > in
> > > > > this scenario as well.
> > > > >
> > > > >
> > > > > On 10/30/2012 03:56 AM, Nicolas Filotto wrote:
> > > > >
> > > > >
> > > > > Do you have new results to provide? Especially on PGSQL?
> > > > >
> > > > >
> > > > > On Mon, Oct 29, 2012 at 9:26 PM, Nick Scavelli <
> > > > > nscavell(a)redhat.com
> > > > > > wrote:
> > > > >
> > > > >
> > > > >
> > > > > Some quick tests suggest there seems to be some improvement
> > > > > however
> > > > > the difference between 1.14 and 1.15 is still there. And it
> > > > > still
> > > > > seems to be the time that is spent in lucene updating the index
> > > > > (IndexWriter#addDocument). With the new jar it seems to
> > > > > indicate
> > > > > a
> > > > > little less time is being spent here, but no where close to the
> > > > > time
> > > > > spent in JCR1.14.
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > On 10/26/2012 11:31 AM, Nicolas Filotto wrote:
> > > > >
> > > > >
> > > > > My mistake jcr.core is not enough you need core.database too,
> > > > > you
> > > > > can
> > > > > get the latest versions from here
> > > > >
>
http://repository.exoplatform.org/content/groups/public/org/exoplatform/j...
> > > > > and here
> > > > >
>
http://repository.exoplatform.org/content/groups/public/org/exoplatform/c...
> > > > >
> > > > >
> > > > > On Fri, Oct 26, 2012 at 7:37 AM, Michal Vanco <
> > > > > mvanco(a)redhat.com
> > > > > >
> > > > > wrote:
> > > > >
> > > > >
> > > > > Same for me with mysql when I was about to try initialization
> > > > > of
> > > > > manyPages in portal with upgraded jcr core. Log is here:
> > > > >
http://pastebin.com/yqTXEbs5
> > > > >
> > > > > Michal
> > > > >
> > > > >
> > > > > ----- Original Message -----
> > > > > > From: "Matt Wringe" < mwringe(a)redhat.com >
> > > > >
> > > > >
> > > > > > To: "Nick Scavelli" < nscavell(a)redhat.com >
> > > > > > Cc: "gatein-dev" < gatein-dev(a)lists.jboss.org
>
> > > > > > Sent: Thursday, October 25, 2012 10:12:20 PM
> > > > > > Subject: Re: [gatein-dev] Performance degradation between
JCR
> > > > > > 1.14
> > > > > > and 1.15?
> > > > > >
> > > > > > same with mysql
> > > > > >
> > > > > > On Thu 25 Oct 2012 04:05:52 PM EDT, Nick Scavelli wrote:
> > > > > > > Using both v1 and v2 I get errors in postgres. Full
> > > > > > > stacktrace
> > > > > > > here
> > > > > > >
http://pastebin.com/BFAMW85M
> > > > > > >
> > > > > > > On 10/25/2012 08:43 AM, Nicolas Filotto wrote:
> > > > > > >> Here is the latest version
> > > > > > >>
>
http://dl.dropbox.com/u/20836627/exo.jcr.component.core-1.15.0-CR2-SNAPSH...
> > > > > > >>
> > > > > > >> On Wed, Oct 24, 2012 at 9:57 PM, Matt Wringe <
> > > > > > >> mwringe(a)redhat.com
> > > > > > >> <mailto: mwringe(a)redhat.com >> wrote:
> > > > > > >>
> > > > > > >> On 24/10/12 03:33 PM, Nicolas Filotto wrote:
> > > > > > >>> Could you please retry with this jar?
> > > > > > >>
> > > > > > >> I tested it quickly with the updated jar, and it
seems to
> > > > > > >> be
> > > > > > >> much
> > > > > > >> closer to the 1.14 results in my test when dealing
with
> > > > > > >> 500
> > > > > > >> pages. I will let Nick comment if it fixes the
issues he
> > > > > > >> was
> > > > > > >> seeing with import/export of sites.
> > > > > > >>
> > > > > > >>
> > > > > > >>>
> > > > > > >>> On Mon, Oct 22, 2012 at 2:51 PM, Nick Scavelli
> > > > > > >>> < nscavell(a)redhat.com <mailto:
nscavell(a)redhat.com >>
> > > > > > >>> wrote:
> > > > > > >>>
> > > > > > >>>
> > > > > > >>> Here are results of some quick tests I ran on
Friday
> > > > > > >>> using
> > > > > > >>> PostgreSQL. Note this is total time spent
saving session
> > > > > > >>> after 500 pages, not time spent indexing.
> > > > > > >>>
> > > > > > >>> 1.15: 112.916s, 113.433s, 110.835s
> > > > > > >>>
> > > > > > >>> 1.14: 69.711s, 81.118s, 81.680s
> > > > > > >>>
> > > > > > >>> For 400 pages (only ran this test once)
> > > > > > >>> 1.15: 61.503s
> > > > > > >>> 1.14: 57.856s
> > > > > > >>>
> > > > > > >>> And the less pages the more closely 1.15 seems
to perform
> > > > > > >>> close to 1.14. This is a very low sample size,
and my
> > > > > > >>> tests
> > > > > > >>> might not best paint an accurate picture, but
it does
> > > > > > >>> seem
> > > > > > >>> that 1.14 consistently out performs 1.15. So
take it as
> > > > > > >>> you
> > > > > > >>> will. I'm also curious what performance is
like if I
> > > > > > >>> would
> > > > > > >>> save the session after every page, instead of
one large
> > > > > > >>> commit of 500 pages.
> > > > > > >>>
> > > > > > >>> - Nick
> > > > > > >>>
> > > > > > >>
> > > > > > >
> > > > >
> > > > >
> > > > > > _______________________________________________
> > > > > > gatein-dev mailing list
> > > > > > gatein-dev(a)lists.jboss.org
> > > > > >
https://lists.jboss.org/mailman/listinfo/gatein-dev
> > > > > >
> > > > > _______________________________________________
> > > > > gatein-dev mailing list
> > > > > gatein-dev(a)lists.jboss.org
> > > > >
https://lists.jboss.org/mailman/listinfo/gatein-dev
> > > > >
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > gatein-dev mailing list gatein-dev(a)lists.jboss.org
> > > > >
https://lists.jboss.org/mailman/listinfo/gatein-dev
> > > > >
> > > > >
> > > > >
> > > >
> > >
> > >
>