Hi, Probably, data are bufferized and for small quantity of data (less than the buffer size) a timeout (near 100ms) is raised to alert the listener that buffer can be read !!!
> -----Message d'origine----- > De : Bertrand Goupil [mailto:[EMAIL PROTECTED]] > Envoy� : mercredi 10 octobre 2001 12:33 > � : [EMAIL PROTECTED] > Objet : Soap benchmark, String performances on weblogic 6.1 > > > Hello every body > > We are preparing a benchmark to compare SOAP and RMI on different > application server. One of our test is like an echo service, we put a > string to serveur and the server return this string on the client. > We have two tests one � 100 characters and a second � 4000 > charact�rs running on weblogic 6.1 with Apache SOAP 2.2. > > The results are: > (VU for Virtual User) > Tests serialize round-trip deserialize total (ms) > 100 1 VU 3,5 171,9 1,7 177,1 > 100 10VU 11,4 193,6 2 ,9 207,9 > 100 50 VU 22,9 332,5 5,4 350,8 > > 4000 1 VU 3,5 19,6 2,0 25,1 > 4000 10 VU 18,2 76,8 6,2 101,2 > 4000 50 VU 24,4 261,8 8,8 295,0 > > The problem is that the 4000 characters test is faster than the 100. > I Made other test and I saw that before 1000 characters the round-trip > is 180 ms and after 1000 characters the round-trip is 16 ms. > > I catch network trafic for this two tests and there is a delta time at > 120 ms after the header packet and a insignificant delat time for more > than 1000 characters. > > So My question is why there is this delta time with less than 1000 > characters or why the 4000 characters tests is faster than 100 > characters. > I check my code and every thing is well > > Thanks for your response. > > Bertrand > >
