On 10.02.2004 10:09, MAHE Vincent FTRD/DIH/REN wrote:

I am running performance tests on a webapp using Cocoon 2.1.3 on Resin 2.1.9 with the IBM JVM.
I have installed the profiler which told me that most of time is spent on the xslt tranform in my pipeline (the setup time, not the processing time)
For the moment I'm using the default xalan XSLT processor.
So I have changed the use-store to "true" (the default value is false, I don't know why), and it increases response time a lot.
When this parameter is set to "false", I can only have 22 simultaneous virtual users (if I want to have a web page in less than one second).
When this parameter is set to "true", I can have 100 simultaneous virtual users (with pages produces in less than 2 seconds).
Here are my questions :
- what is the exact role of the use-store parameter ?
- what are the advantages/drawbacks of using this parameter ? Why is it set by default to "false" in Cocoon 2.1.3 ?
- what is the best xslt processor to use (Xalan, XSLTC, Saxon, ...). Are there any bench results about that, using Cocoon ?
It makes a lot of questions but It seems that xslt is the major performance point to tune. If there could be more details on the Cocoon documentation about that, it would be very helpfull for Cocoon starters like me.

Some very long time ago there was a memory leak issue AFAIK. The issue of use-store has been often discussed in the last year on the developers list, so you should find mails in the archives, e.g.


http://marc.theaimsgroup.com/?t=104696353000002&r=1&w=4
http://marc.theaimsgroup.com/?t=104704747000010&r=1&w=4
http://marc.theaimsgroup.com/?t=106441161500004&r=1&w=4
http://marc.theaimsgroup.com/?t=106737246000001&r=1&w=4

Joerg

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to