I agree as well. And even when scaling out, if you use many machines running lots of threads all trying to report back to a central controlling instance of meter it's possible to overwhelm them all. I did this, without having adjusted any heap sizes, by trying to have 10 machines run hundreds of threads each. I exhausted heap space on all of them. Of course part of it could have also been my test design, but to really generate big load and collect meaningful, measurable, reliable results you've got to scale out beyond one machine.
-- Charles H. Baker, Systems Programmer II CCIT/CSO Don't see me in my cubicle? Try AIM: rascharles GTalk: [email protected] [email protected] | 864.656.8069 | 864.990.1297 Give a man a fish and you feed him for a day. Teach a man to fish and you feed him for a lifetime. Chinese Proverb On Mar 1, 2012, at 11:50 AM, Anthony Johnson wrote: > I agree with everyone else about the 300-400 thread count being a good > limit for JMeter. I have taken JMeter above those values, but you are > playing in Out of Memory territory. > > Even if you get the test to execute, 6,000 threads are simply going to > compete for resources and yield unreliable results on a single > machine. > > You either need to re-think your test plan to use less threads or > scale out the test to use multiple machines. > > Thanks, > > Anthony > > On Thu, Mar 1, 2012 at 11:38 AM, Flavio Cysne <[email protected]> wrote: >> We used to execute distributed testing using JMeter agent (jmeter -s) >> loading almost 400 threads each. On that case we were using HEAP >> with 1,5GB, under a virtualized Linux CentOS with 2 cores processors and >> 3GB RAM box. >> >> Keep in mind that for better results the cpu/memory usage don't be higher >> than 80%. It can cause anomalies to test results in some cases. >> >> Hope it helps you. >> Flávio Cysne >> >> 2012/3/1 ZK <[email protected]> >> >>> Hi, >>> 6000 threads seems like a lot to me for 1 machine >>> I normally dont go above 300 threads per load injector >>> >>> ZK >>> >>> -- >>> View this message in context: >>> http://jmeter.512774.n5.nabble.com/J-Meter-fails-when-thread-count-is-increased-tp5527549p5527736.html >>> Sent from the JMeter - User mailing list archive at Nabble.com. >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: [email protected] >>> For additional commands, e-mail: [email protected] >>> >>> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
