Thank you for the advice. I've not much experience profiling with java,
I didn't know that the time invoking counting is counted in the result.
That doesn't seem very smart to me either.
Johan Compagner wrote:
all high accurate settings i know of (also yourkit) are influencing the
test itself!
You shouldn't use method tracing/invoke counting with yourkit either..
(only if you really need invoke counting)
Because the time it takes to record those things are also in the trace
itself (pretty stupid)
method sampling of yourkit is pretty much close to what you feel in the
real world.
johan
Matej Knopp wrote:
Hmm.. Using YourKit Java Profiler, wicket really is only marginaly
slower. I wonder why JProfiler makes such a difference (on the highest
accurate settings).
Matej Knopp wrote:
Thank you.
The results I get are mixed. When editing acustomer entry and saving
it, the performance is comparable.
But deleting the customer, wicket is still almost twice as slow than
tapestry.
Have you tried to do nothing but deleting like 10 customers? I'm
under an impression that rendering listview with many items like this
is the slowest part.
Chris Turner wrote:
I actually didn't really polish my setup. I just grabbed some stuff
I did on a previous project which made use of the Spring Framework
support for log4j configuration. Basically:
1) Copy spring framework jar to your webapp WEB-INF/lib directory
2) Add the following to your web.xml:
<context-param>
<param-name>log4jConfigLocation</param-name>
<param-value>/WEB-INF/log4j.xml</param-value>
</context-param>
<servlet>
<servlet-name>log4jConfig</servlet-name>
<servlet-class>org.springframework.web.util.Log4jConfigServlet</servlet-class>
<load-on-startup>1</load-on-startup>
</servlet>
3) Create a log4j.xml file:
<log4j:configuration xmlns:log4j="http://jakarta.apache.org/log4j/";
debug="false">
<appender name="CONSOLE" class="org.apache.log4j.ConsoleAppender">
<param name="Target" value="System.out"/>
<param name="Threshold" value="INFO"/>
<layout class="org.apache.log4j.PatternLayout">
<!-- The default pattern: Date Priority [Category]
Message\n -->
<param name="ConversionPattern" value="%d{ABSOLUTE} %-5p
[%c{1}] %m%n"/>
</layout>
</appender>
<category name="">
<priority value="INFO"/>
</category>
<root>
<appender-ref ref="CONSOLE"/>
</root>
</log4j:configuration>
Matej Knopp wrote:
Could you please post somewhere the war file with logging disabled?
I tried adding log4j.properties with the following content
log4j.threshold=info
log4j.debug=false
into the wicket-benchmark class path but it helped only a little.
It seems that I'm doing something wrong.
Thank you.
Chris Turner wrote:
I believe that all of the logging statements are currently
surrounded by the if check - I could not find any that were not.
However, if you don't provide a logging configuration then the
call to log.isDebugEnabled() always returns TRUE. In my opinion
this has always been a fault of commons logging/log4j. In my
opinion if you don't configure debug logging to be explicitly on
then it should ALWAYS be FALSE.
The problem as I see it is that we will always end up with
problems like this where someone evaluating wicket just knocks up
a test application, does a quick performance comparison with their
existing framework, finds our performance very poor (because debug
is on by default) and throws wicket away without investigating
further.
Therefore we must either find a way to ensure that people cannot
easily build a wicket application with debug logging left on by
default, or if we can't do this then we must minimise the
performance impact of them forgetting to turn it off!
regards,
Chris
Johan Compagner wrote:
If you are in loggin debug mode then
log.debug("Begin render " + this);
is fine by me
BUT
it should always be checked first:
if(log.isDebugEnabled())
{
log.debug("Begin render " + this);
}
Chris Turner wrote:
Confirmed. Turn off the logging (initialise log4j and set level
for wicket to INFO or greater) and Wicket is only about 4%
slower than the Tapestry application. I suspect with a little
effort we could soon tune out this small amount :) Probably
best to wait until the 1.1 code base is complete to do this as
refactorings are notorious for undoing previous performance
optimizations.
However, the question is therefore what to do with logging: One
option is to ship Wicket example applications with correctly
configured log4j to turn logging off by default. This would also
require VERY clear documentation to explain that Wicket logging
should be turned off by default due to the application
performance hit of having it switched on. Alternatively, we
should just go through the code and remove all of the debug
render or at least refactor the statements of the form:
log.debug("Begin render " + this); which are very expensive
because they do a toString() on the object.
Any other thoughts?
regards,
Chris
Johan Compagner wrote:
if that really is the case then we should turn logging default
off??
and supply a right properties file?
johan
Chris Turner wrote:
One word: *** LOGGING ***
Looking through the profile in detail it appears that the
benchmark application has no logging configured so by default
logging is ON. Many of the preformance bottlenecks seem to be
in building the strings for logging. I'm going to figure out
how to turn off logging and then I'll try the profiling again.
regards,
Chris
Chris Turner wrote:
I have just run the two examples through a profiler (YourKit
Java Profiler) and I can confirm that the Tapestry example is
just over twice as fast as the Wicket one. The amount of
difference is variable, with the biggest difference being for
the page that renders the list view.
Taking a quick look through the profile trace for the Wicker
code there does not seem to be any one area where there is a
particular performance bottleneck. However, most of the
largest areas of execution time resolve down to one of two
cases:
- Methods that do lots of string manipulation (in particular
toString and StringBuffer.append stuff)
- Methods that invoke Ognl
Expanded report showing main bottlenecks as identified by
YourKit: http://www.skipoles.co.uk/wicket-benchmark.zip
I'll investigate a bit further.....
regards,
Chris
Eelco Hillenius wrote:
That's a scarry difference. We should take a good look at
that and see what causes it. Personally, I wouldn't expect
our 'statefullness' to be the real problem here; as 'heavy'
as it sounds to create new objects, this is something Java
should be very good at. We should use a profiler to find out.
Eelco
Phil Kulak wrote:
I just did some quick testing using the two wars. I did 100
page loads
with a brand new session on each load and Tapestry was
about twice as
fast for the list view page and the much smaller edit page.
I wonder
what would happen if I made the Tapestry app stateful as
well. That
may be a more fair comparison.
On 7/10/05, Phil Kulak <[EMAIL PROTECTED]> wrote:
Arg, okay, that was to much to post. Here's the origonal
message:
Here are the two wars. Let me know if I should change the
functionality at all. Also, notice that I didn't try very
hard to
optimize the wicket application. For example, I don't use
optimizeItemRemoval = true and then manually removeAll()
on deletes. I
could have done that, but there are probably corrosponding
optimizations for Tapestry that I just don't know about
because I
don't use it. Also, the Tapestry app is totally stateless,
so that may
give it an edge, but I dunno for sure.
On 7/10/05, Phil Kulak <[EMAIL PROTECTED]> wrote:
Okay, I just realized that it was a bit dumb to post 5
megs of files
to mailing list. Sorry about that. Here are some URLs:
http://gladstone.uoregon.edu/~pkulak/tapestry-benchmark.war
http://gladstone.uoregon.edu/~pkulak/wicket-benchmark.war
On 7/10/05, Phil Kulak <[EMAIL PROTECTED]> wrote:
Here are the two wars. Let me know if I should change the
functionality at all. Also, notice that I didn't try
very hard to
optimize the wicket application. For example, I don't use
optimizeItemRemoval = true and then manually removeAll()
on deletes. I
could have done that, but there are probably corrosponding
optimizations for Tapestry that I just don't know about
because I
don't use it. Also, the Tapestry app is totally
stateless, so that may
give it an edge, but I dunno for sure.
-Phil
On 7/10/05, Phil Kulak <[EMAIL PROTECTED]> wrote:
If there's just one DAO, then I can store the data in a
Map and not
worry about HSQL or anything like that. Actually, I
just wrote it in
Wicket (it's absolutly frickin' amazing how FEW lines
it took) and
I'll get it cloned in Tapestry tomorrow and post up
both wars.
Basicaly, it's just a bare-bones version of your
CD-App. I didn't want
to make it too complicated because I'm not super
efficiant with
Tapestry and I didn't want that part to take all day.
-Phil
On 7/10/05, Eelco Hillenius
<[EMAIL PROTECTED]> wrote:
Phil Kulak wrote:
Unless the dynamic images are actually created on the
fly, Wicket's
gonna get slaughtered there since Tapestry will
actually copy static
images from the jar into a public webserver directory.
I think Wicket handles the static images pretty
efficiently. It would be
interesting to see if there is any performance
difference with Tapestry,
but I think not. So Tapestry actually copies the
resources to the webapp
directory? A potential problem with that is that you
can't be sure that
you have write rights in your webapp dir. You can't
even be sure if
there is a filesystem dir at all (might still be a
.war, or potentially
some other schema).
Maybe a list view of 50 items with edit and delete
options? I could
make a single, thread-safe DAO, and then if you could
find a way to
hit the page a bunch of times with concurrent edits
and deletes, it
would be a nice muli-user test.
Why should you have just one DAO? Why not an instance
per request or
session?
That's my idea, but I have no idea how you plan to
test it, so
something else may be more appropriate.
I think it would be nice to have a special performance
test page. Where
should we put it though?
Eelco
On 7/9/05, Johan Compagner <[EMAIL PROTECTED]> wrote:
If i get 2 wars of an app that does some big
rendering (like listviews,
dynamic images and that kind of thing)
So i don't think hangman will do it really it is a
bit to small.
Don't know if we really have to test a database.
That i don't like. It
should be plain pojo's
that are in mem created for a list/detail/edit view.
What kind of functionality would be a good test?
johan
Phil Kulak wrote:
Is this offer still available? Don't we already have
a working Hangman
app in both frameworks? If that won't cut it, I'd be
glad to write a
Tapestry clone of a Wicket app. It would give me a
chance to get
better acquainted with Tapestry 4.
On 6/18/05, Johan Compagner <[EMAIL PROTECTED]>
wrote:
then we need 2 the same apps with the same feature
set in both frameworks.
If somebody is willing to build that and supply me
with the 2 wars then
i will do a performance test
with yourekit..
johan
叶卫国 wrote:
Does anybody compare it with tapestry?
-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux
Migration Strategies
from IBM. Find simple to follow Roadmaps,
straightforward articles,
informative Webcasts and more! Get everything you
need to get up to
speed, fast.
http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux
Migration Strategies
from IBM. Find simple to follow
Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you
need to get up to
speed, fast.
http://ads.osdn.com/?ad_idt77&alloc_id492&opclick
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
N琀S^甸殜X�'矈辵辑呧�Z+a妤伶鉂j歗�)�&閞讍⿵
?跽﹦
5瀶{獾歙黑 茩hazV瓃薭澺殨�+y┹v妤偠﹩',电!瀴h�&瑍
窞 、 贽 介韱�-y烛 �
┹5R
璀�"沧+"�m��鹈ir壙倧莨﹑y抚j耽rG谦櫒x%娝V壣
峨 甔 � (悍 ~娻zw瓎踚��鍔薼矉玵玷�咤娝l⺋)撸
�"rG谦
N�HS^�隊X���'���u�����2��Z+a���❪�j�^�)�&�r׆��
?�թ�
5��{�쨺� ƙh��azV�z�b�ۚ��+y��v楂���',��!��h�&��
�~���w���� 톋-y�� �
��5R
��"��+"�m����ir����ݹ�py��j��rG��ǫ����x%��V�� �� �X���(�� ~��zw���i����l���q���z���l�X��)ߣ�"rG��ǫ
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With
Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore
the latest in dual
core and dual graphics technology at this free one
hour event hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
N�HS^�隊X���'���u�����2��Z+a���❪�j�^�)�&�r׆��
?�թ�
5��{�쨺� ƙh��azV�z�b�ۚ��+y��v楂���',��!��h�&��
�~���w���� 톋-y�� �
��5R
��"��+"�m����ir����ݹ�py��j��rG��ǫ����x%��V�� �� �X���(�� ~��zw���i����l���q���z���l�X��)ߣ�"rG��ǫ
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the
latest in dual
core and dual graphics technology at this free one hour
event hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the
latest in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the
latest in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the
latest in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest
in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit
http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!'
webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest
in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest
in dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in
dual
core and dual graphics technology at this free one hour event
hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in
dual
core and dual graphics technology at this free one hour event hosted
by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted
by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted
by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar
happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
Wicket-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-user
