cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From TREGAN Fabien <>
Subject RE: Performance of XSLTC
Date Thu, 11 Jul 2002 06:07:54 GMT
I have a pipeline with 1 generator, 1 castor (non cachable) transformer, 2
xsl transformation and 1 xml serialization. I tryed to run it with both
xalan and xsltc, and the benefit (mesured by JMeter with 5 and 50 concurrent
threads) gave me a around 10% gain on the whole pipeline (that is less than
15% for the transformation, since the xslt transformer take nearly all the

Did you really have a 50% benefit ?

>-----Message d'origine-----
>De: Berin Loritsch []
>Date: jeudi 11 juillet 2002 00:54
>Objet: RE: Performance of XSLTC
>> From: Reinhard Poetz [] 
>> Hi,
>> Today I tried to find out how much faster XSLTC compared to 
>> Xalan is. My
>> result: there is hardly any difference in performance (maybe 
>> XSLTC is slightly faster).
>> My test case:
>> -------------
>> I used a pipline with two xslt transformation steps. The 
>> first time I used the standard transformer (XSLTC) and then I 
>> changed the type to "xalan". I'm using Cocoon2.1dev - checked 
>> it out today.
>> I got the results (measurements) from the profiler.
>> I know that this was not a valid test series and only a first 
>> impression - but ...
>> ... now I have some questions:
>> ------------------------------
>> - Am I doing something wrong?
>> - Do I have to make a real load test to get a difference in 
>> performance?
>> - Did anybody get other/similar results?
>For testing the performance differences between the implementations,
>you must not mix and match implementations.  It should be all
>Xalan or all XSLTC.  Also, it is very important to TURN OFF CACHING!!!
>You will not be measuring differences in the transformation engines
>if caching is on.
>With two identical pipelines, with the exception of XSLTC vs. Xalan,
>measure the time it takes for the whole request.  The difference
>between the times will be the result you are looking for.
>Do average 10 samples, dropping the high and the low samples.  That
>will help you determine the real speed difference.
>A) G->Tc->S
>B) G->Tx->S
>If pipeline A is 30ms and pipeline B is 60ms on average, then the
>benefit of using XSLTC is 50%
>To unsubscribe, e-mail:
>For additional commands, email:

To unsubscribe, e-mail:
For additional commands, email:

View raw message