On 12 September 2011 00:59, Oliver Lloyd <oliver_ll...@hotmail.com> wrote: > I'm thinking of running multiple jmeter processes on a single machine, up to > 20. My testing so far has not highlighted any issues with this in principle > - I can execute the same jmx file over multiple java processes and I am > careful to write to separate jtl files. > > The reason I'm taking this approach is to meet a requirement for a test > simulating a very large number of concurrent connections for an application > using long-polling. My test is extremely simple: one GET request which is > held open by the server for 15 seconds and then immediately resent, that's > it. Running this test does not create a heavy load on the machine JMeter > runs on - the throughput is low - but I am not able to run more than about > 1000 threads within a single JM process (I get OOM errors past that). I need > quite a lot more than 1000 so by vertically scaling multiple JM processes I > am aiming to have each physical machine hold open about 10-20,000 > connections and then scale these out horizontally to get the load I need. > > Doable?
It sounds like you won't run out of CPU, and I assume you have enough physical memory to support all the JVM instances. However, depending on the OS, you may find it tricky to support so many connections. You may need to tweak the OS settings. > ----- > http://www.http503.com/ > -- > View this message in context: > http://jmeter.512774.n5.nabble.com/Vertically-Scaling-JMeter-tp4792729p4792729.html > Sent from the JMeter - User mailing list archive at Nabble.com. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: jmeter-user-unsubscr...@jakarta.apache.org > For additional commands, e-mail: jmeter-user-h...@jakarta.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: jmeter-user-unsubscr...@jakarta.apache.org For additional commands, e-mail: jmeter-user-h...@jakarta.apache.org
