to increase.
Any thoughts anyone. Looking specifically for the conclusion from the OP.
--
View this message in context:
http://old.nabble.com/-Trinidad--Trinidad-consuming-80-90--CPU-tp27068136p28750989.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
to increase.
Any thoughts anyone. Looking specifically for the conclusion from the OP.
--
View this message in context:
http://old.nabble.com/-Trinidad--Trinidad-consuming-80-90--CPU-tp27068136p28750991.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
2010 09:22
An: users@myfaces.apache.org
Betreff: Re: [Trinidad] Trinidad consuming 80-90% CPU
Were any conclusions reached relating to this issue. I have been drafted
in
for an emergency issue that appears on the surface very similar.
The tech stack is near identical - WAS 6.1
-consuming-80-90--CPU-tp27068136p28751130.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
That's put an end to my search for performance fixes in a later 1.0.x
release then. How can you be on 1.0.13? That isn't release? Perhaps you
are
actually on 1.2.13 as this would allign with your faces 1.2
implementation.
Yes, that is correct - it's 1.2.13
That then begs the question of
--Trinidad-consuming-80-90--CPU-tp27068136p28751922.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
.* packages
WAS: 22869
Tomcat: 22092
Which for arguments sake are identica. But, we have 24576 unaccounted for in
WAS as an overhead that is currently being filtered.
--
View this message in context:
http://old.nabble.com/-Trinidad--Trinidad-consuming-80-90--CPU-tp27068136p28752758.html
Sent
-consuming-80-90--CPU-tp27068136p28752812.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
://old.nabble.com/-Trinidad--Trinidad-consuming-80-90--CPU-tp27068136p28752966.html
Sent from the MyFaces - Users mailing list archive at Nabble.com.
What are you seeing in the task manager when you send a request? What
percentage of the CPU is indicated as being used there? I have used JProbe
before and the the results are not that easy to interpret. Much of my
development occurs on a Windows machine with 2Gb of RAM. I have never seen
the
On Thu, Jan 14, 2010 at 9:45 PM, Ravi Kapoor ravikapoor...@gmail.com wrote:
Matthias, I did not hear back from the company you recommended. Can you
check with them?
he told me he would write to you back on Thursday/Friday; let me double check...
-Matthias
Richard, after a painful 2 days, I
Hey Ravi,
First you need to connect to your server. If that's done, depending on
the type of connection, profiling becomes possible (AFAIK local
applications always work, but JMX based connections don't).
If profiling is possible, you should see a Profiler tab. In the
Profiler tab, on the right
Jan-Kees, thanks for this info. I got the settings in order. But I am having
another problem.
When I launch the server from within RAD, VisualVM is not able to detect the
process at all. Can you tell me what do I need to do so VisualVM can detect
my server so that I can get profiling information
Hrm, it looks like the IBM JVM doesn't support VisualVM. See:
http://www.ibm.com/developerworks/forums/thread.jspa?messageID=14210263
Sorry. It would be nice if it worked.
You're probably not able to run your app on i.e. Tomcat, JBoss or
Glassfish, just for testing? If you are, things are easier.
I found that link as well 10 seconds ago and was going to email you :). Yes
you are right, I cannot run this app in tomcat, this is enterprise app with
tons of dependency on websphere.
Anyways, how can I get onsite support? Who provides such support? Apache?
Can you send me contact info or a
Apache officially doesn't provide support, see:
http://httpd.apache.org/docs/1.3/misc/FAQ.html#support
But you can try to contact a commercial company that does provide support.
It's an issue Apache can't really help you with.
Depending on where you're located, it shouldn't be very hard to find
I contacted Ravi already offline
-M
On Tue, Jan 12, 2010 at 7:59 PM, Jan-Kees van Andel
jankeesvanan...@gmail.com wrote:
Apache officially doesn't provide support, see:
http://httpd.apache.org/docs/1.3/misc/FAQ.html#support
But you can try to contact a commercial company that does provide
One suggestion that I would make is to run the Trinidad example application
and examine the behavior.
-Richard
On Tue, Jan 12, 2010 at 11:00 AM, Matthias Wessendorf mat...@apache.orgwrote:
I contacted Ravi already offline
-M
On Tue, Jan 12, 2010 at 7:59 PM, Jan-Kees van Andel
That is an excellent suggestion Richard. Can you point me to the application
you mentioned (I am on trinidad 1.0.7)
Thanks
On Tue, Jan 12, 2010 at 4:58 PM, Richard Yee richard.k@gmail.comwrote:
One suggestion that I would make is to run the Trinidad example application
and examine the
http://myfaces.apache.org/trinidad/download.html
trinidad-1.0.11-example.ziphttp://www.apache.org/dyn/closer.cgi/myfaces/binaries/trinidad-1.0.11-example.zip
I'm not sure if the 1.0.11 version has any incompatibilities with 1.0.7. If
you can't find one that works with 1.0.7, email me offline and
Ravi,
spoke to a guy that does performance testing/improvement for Oracle
Applications. He said that there is some % CPU in Trindad but I would
not give it more them 20%. The heavy hitters is getClientId (Blake -
see dev@ thread - is doing some optimization there).
Now if getProperty is some el
Matthias,
If the issue is in bean, it should show up in my analysis. Also
getProperty is only 40% CPU, there is additional 45% cpu consumed by
rest of the trinidad classes totaling 85% total CPU, all within
org.apache.myfaces.* classes
Ravi
Matthias Wessendorf wrote:
Ravi,
spoke to a
Another thing, most of our EL expressions are one of the following types
#{bean.active}
or
#{bean.get['memid']}
Parsing of these expressions probably happens within org.apache.myfaces.*
classes. These are fairly basic EL expressions and should not be taking much
time. We know, the final getter
Maybe it happens when accessing the value from the Map with
#{bean.get['memid']}, because the Map is not properly synchronized, thus
its internal structure is broken and thus it is running in infinite loops.
Are your resources properly synchronized?
Just a guess in the blue...
Regards,
Jakob
I am thinking if that was the case, I would see high CPU in
java.util.HashMap instead of org.apache.faces.*
If you disagree, please explain and I can try using HashTable to store data
and get fresh numbers.
On Mon, Jan 11, 2010 at 10:00 AM, Jakob Korherr jakob.korh...@gmail.comwrote:
Maybe it
Ravi,
We have load tested our Trinidad application with up to 800,000 page
loads/hr. and the application handled it fine. We have also simulated up to
190 concurrent users. We are using MyFaces 1.1.5 and Trinidad 1.0.5. We are
using the Oracle Application Server 10GR3 and are running on Linux.
Hi Ravi,
Just a stupid question. Are your beans extremely big? Since FacesBean
doesn't do much locking/synchronizing (none if I'm correctly), the
only reason I can imagine it to eat CPU cycles is because of its
size...
Also, is it possible for you to do a profiling run using VisualVM?
I've been
Thanks Richard for the numbers. Can you also mention no of servers used, CPU
and memory details of each server?
Regards
Ravi
On Mon, Jan 11, 2010 at 3:00 PM, Richard Yee richard.k@gmail.comwrote:
Ravi,
We have load tested our Trinidad application with up to 800,000 page
loads/hr. and the
Jakob,
The beans are not big. Also these tests have been performed under a single
user to measure the CPU timing. So I doubt locking/synchronization is an
issue.
I would think that if there is synchronization, then the threads could be
waiting for long time, but lack of synchronization can only
Ravi,
If you are running on Windows, what does the TaskManager show as the CPU
utilization? How much memory is also being used? 2Gb is not much memory for
a production application. I have that much on my desktop.
-Richard
On Mon, Jan 11, 2010 at 1:05 PM, Ravi Kapoor ravikapoor...@gmail.comwrote:
The memory 2GB is not much for a server but even that much is not being used
(due to CPU constraint).Max memory used is about 1.1GB or so when CPU has
hit 100%.
The numbers I posted are from my desktop (which has 4GB memory) as it is
easier to profile CPU on local desktop. The CPU usage is fairly
Our servers are made by Sun and have 4 AMD Opteron processessors with 16GB
RAM. We are running in a cluster of 3 servers.
-Richard
On Mon, Jan 11, 2010 at 12:54 PM, Ravi Kapoor ravikapoor...@gmail.comwrote:
Thanks Richard for the numbers. Can you also mention no of servers used,
CPU
and
Jakob,
I did install VisualVM but it is not capturing org.apache.* classes or
com.mycompany.* classes. I did not find a setting on how to enable capturing
data for all the classes I want. Can you tell me where to specity the
packages for which I want to capture CPU usage?
Vinay
On Mon, Jan 11,
Ravi,
To your question: Can you explain little bit how lack of synchronization can
add to CPU?
I don't know exactly, but maybe it is possible that the internal structure
of a Map (e.g. HashMap) can be destroyed when there are multiple change
operations (put, remove, clear) at the same time,
Oh I see what you are saying. We only create and populate the hashmap once,
we never clear or delete from it. So that endless loop is not the problem.
Also in that case the CPU wouldnt show up for org.apache.* classes.
I guess you are looking for how to specify packages in VisualVM. This would
Richard,
So basically your system can handle 18.5 pages/CPU/second. I can only
wish we could get that kind of throughput. We can only get 2
page/CPU/second or so.
If you can get that kind of performance, I wonder what is wrong with our
environment. I think I need a good night sleep and
Hello Ravi,
I wonder what our last release for JSF 1.1 (1.0.11) does?
Not sure, perhaps you may also try the JSF 1.2 version ? (1.2.12)
The JSF 1.2 version is the one that is best supported, these days.
Trinidad 2.0 is now in alpha stage, and I can understand that you
don't want to update on
Matthias, I think websphere 6.1 does not support JSF 1.2. I will
doublecheck, let me know if this is incorrect. This mans I cannot try
trinidad version 1.2.12
I will try out 1.0.11 release, but that is a minor release update and I
seriously doubt if it will fix such a performance issue.
Hey Ravi,
Looking at your JProbe screenshots for the second time, I think you're
misinterpreting the graphs (but I haven't used JProbe before, so I
might be mistaking ;-) ).
In your first screenshot (upper left corner) you can see the total
time the getProperty method takes.
This includes its
Hi Jan-Kees,
You are right, the getProperty method is only taking 2K units. However if I
dig deeper, I find that most of the cumulative time is being spent within
Trinidad classes. The final call to java getters consumes negligible time.
I was unable to create thread structure like you showed
I don't know. I'm of the camp that if the CPU time is available, use
it. That said, is this load consistant or are you just testing an
initial hit of each page.
Sent from my iPhone
On Jan 8, 2010, at 11:25 PM, Ravi ravikapoor...@gmail.com wrote:
Hi Jan-Kees,
Now that I am reading your
Scott, we do not have CPUs available. The time trinidad is consuming is
supposed to be doing some other work. Hence this is costing us real
dollars and hence our time and effort to resolve this.
This is not initial hit of page. I always ignore the first hit on all
pages, I am only measuring
Hi Matthias,
Here are the details:
Server: Websphere 6.1
Trinidad version: 1.0.7 (We cant upgrade to 2.0 until we upgrade websphere
which will happen in due course. Even then if this issue has not been
addressed, the problem may exist in 2.0 as well.)
OS: Windows (Even though I am measuring
Hey,
Is it possible that the getProperty indirectly invokes some expensive
computation? For example, do you have lots of logic inside your
getters?
Regards,
Jan-Kees
2010/1/8 Ravi Kapoor ravikapoor...@gmail.com:
Hi Matthias,
Here are the details:
Server: Websphere 6.1
Trinidad version:
The actual call to getter method is only using 2% CPU. Rest 38% is being
used within trinidad classes.
I am attaching two screenshots to give you more details.
In first screenshot, you can see at the top left corner, total CPU units
taken by getProperty are 32391
getProperty calls
I'm not sure, but I doubt the mailing list supports attachments.
Maybe you could provide a link to some image hosting site?
My first thought, reflection is darn cheap, especially since Java 5
and even more since Java 6. I'm no IBM JVM specialist, but I don't
think there are major differences with
How much physical memory is on your testing machine?
I have a few Trinidad applications in production and don't see any of the
performance issues you are having.
-Richard
On Fri, Jan 8, 2010 at 12:33 PM, Jan-Kees van Andel
jankeesvanan...@gmail.com wrote:
I'm not sure, but I doubt the
This is an enterprise application. I am using tracing not sampling.
The thought that the issue within application and not in trinidad occurred
to me as well. Hence I validated in many ways and everything points to
trinidad as having the issue.
I measured the CPU used within each class and this
I have 2 GB on the machine and it only uses 1 GB.
Can you give details on your environment. Especially trinidad version, CPU
details and how many users per JVM can you handle, what %age of CPU is
consumed by trinidad etc
Regards
Ravi
On Fri, Jan 8, 2010 at 4:08 PM, Richard Yee
Hi Jan-Kees,
Now that I am reading your message again, I do want to answer your
questions in detail. First I agree reflection is cheap, that is why
reflection is not my concern. Time being spent in reflection is almost
negligible compared to time being spent in trinidad classes.
Secondly
Has anybody done performance tests on trinidad application. I have an
application and it appears that it is taking 80-90% of CPU in my
application, thus killing performance.
We ran load tests and our CPU went to 100% usage. At this point we measured
how much time was being taken by each
Hello Ravi,
some more background would be good, e.g. what version of Trinidad etc.
-Matthias
On Thu, Jan 7, 2010 at 11:25 PM, Ravi Kapoor ravikapoor...@gmail.com wrote:
Has anybody done performance tests on trinidad application. I have an
application and it appears that it is taking 80-90% of
52 matches
Mail list logo