RE: Axis Tomcat performance

[Benazech Cédric <Ce...@atosorigin.com>]

starting a java program take a long time because of the starting of the JVM.
That explain your bad results using time command.

If you do 100 calls to a webService in 1 java program, you will see that the
performance will be greatly better than doing 100 start of a java program
that do 1 call.

Cédric


-----Message d'origine-----
De : Rodrigo Serra [mailto:rmserra@fibertel.com.ar]
Envoyé : mardi 30 avril 2002 17:22
À : axis-user@xml.apache.org
Objet : RE: Axis Tomcat performance


I have the same problem but with the Calculator example. The output of
command "time" is:
$ time java samples.userguide.example2.CalcClient -p8080 add 2 5
Got result : 7
real    0m2.426s

The machine is Pentium III 1GZ.
I observe  client java virtual machine use all CPU in contrast with the
tomcat process not consume much cpu.

-----Mensaje original-----
De: Geza.Szocs@nokia.com [mailto:Geza.Szocs@nokia.com] 
Enviado el: Martes, 30 de Abril de 2002 06:01 a.m.
Para: axis-user@xml.apache.org
Asunto: RE: Axis Tomcat performance

> Hi Geza,
> 
> if I set the option to "XXX", the server returns a float 
> directly. Only if 
> I set the option to "IBM", the server get the stock from the 
> real internet.
> Nethertheless, the misc example ,which will send a hardcoded 
> message to the 
> server and returns that message will take about 4 seconds.
> 
> Do you think, the performance will be better on a 
> Windows-machine accessing 
> localhost ?

Well, I dunno. I just wanted to explain the answer given to
previously. Anyway, you are scaring me. I'm just doing some
stuff with Axis, which is pretty performance-critical. 
With Beta 1 I did some benchmarking myself, but the result was
not so bad. I measured well below 1 second. Around 1-5 millisec,
if I remember it well. It was just a simple return-some-string
example. I can't image what could be wrong with your setup, sorry. :(
I just hope that my measurement is closer to the real truth, otherwise
we're in deep sh@t. :)