I got some troubles as soon as i tried to use about 20 threads. This was because i was testing a HTTP1.1 server (silverstream) using jmeter on JDK1.3. The JDK1.3 does not seem to implement the HTTP1.1 protocol correctly. I got around this by using good old jdk1.2 (which implements HTTP1.0) to run jmeter. Now I can get up to fifty threads. After that, weird things seem to happen on the server. I think it has something to do whith the LDAP server silverstream uses for authentication. It seems that this LDAP server likes to take a lot of TCP connections, leading to random ServiceNotAvailable exceptions. But this is ofcourse not a limitation of jmeter.
> I have run with up to 300 threads (on Solaris). I think that I could get > to 500 with some more tweaking, > -----Original Message----- > From: Tom Wang [mailto:[EMAIL PROTECTED]] > Sent: October 11, 2001 3:21 PM > To: Paul Devine; [EMAIL PROTECTED]; [EMAIL PROTECTED] > Subject: RE: Attainable Load > > I would say 100 is more than enough for me to do concurrent user testing. > > > -----Original Message----- > From: Paul Devine [mailto:[EMAIL PROTECTED]] > Sent: Thursday, October 11, 2001 12:16 PM > To: Tom Wang; [EMAIL PROTECTED]; [EMAIL PROTECTED] > Subject: Attainable Load > > I am curious what level of load people have attained with JMeter. I > realize this is a somewhat open ended question depending on many factors. > I'm curious on a realistic attainable number of truly overlapping requests > to a server under test. At some point I assume JMeter will run into > exceptions because the JVM will not give out any more socket connections > due to JVM and/or OS level limits. I've never had problems generating > up to a hundred users in whatever I've tested so far. (With other tools > on Linux I've run into limits which seemed to be JVM/OS related but I > can't recall what those limits were.) > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
