Hi Shumel, I could not really get this resolved completely. I am customizing my own sampler that is much light-weighted and I found a significant improvement in terms of cpu and memory after using it.
Regards, Jacky On Tue, Aug 7, 2012 at 10:48 AM, Shmuel Krakower <[email protected]> wrote: > Hi Jacky, > How did you resolved this? If you did. > > I think it can be helpful to the community to have some more info on how to > enable jmeter on big boxes. > > Best, > Shmuel. > > On Mon, Jul 30, 2012 at 11:32 PM, Flavio Cysne <[email protected]>wrote: > >> With this huge :D box I would try to: >> >> - Use JMeter's SaveService in batch mode to achieve a better performance; >> - Use a high throughtput and low pause GC JVM configuration ( >> >> http://java.sun.com/performance/reference/whitepapers/tuning.html#section4.2 >> ); >> - Use at least 4 JMeter's servers within this machine (-Xms1g -Xmx1g, maybe >> 2g) and run this test in distributed mode. >> >> You could also install libAIO package. I have experienced some GC >> improvements after installing it. >> >> The directive sun.rmi.dgc.client.gcInterval is expressed in milisseconds, >> so I'll vote to change this to a higher value, let's say 60000 (one >> minute), or simply remove it. The default value to JDK6 is 3600000 (one >> hour). - >> >> http://docs.oracle.com/javase/6/docs/technotes/guides/rmi/sunrmiproperties.html >> >> Other references to JVM Tuning: >> http://www.oracle.com/technetwork/java/ergo5-140223.html >> >> Hope it helps you. >> Flávio Cysne >> >> 2012/7/30 unjc email <[email protected]> >> >> > Flavio, my box has 12GB RAM and 16-processors installed with >> > Centos6.2, JDK 1.6 and Jmeter 2.7. >> > >> > >> > Thanks. >> > >> > On Mon, Jul 30, 2012 at 7:39 AM, Flavio Cysne <[email protected]> >> > wrote: >> > > Can you provide us with your box configuration? OS distribution and >> > > version, JDK version, JMeter version, Hardware spec. >> > > >> > > Try with other configurations of GC like -Xincgc, -XX:UseParallelOldGC, >> > > -XX:UseConcMarkSweepGC, etc (one at a time) and compare the results. >> > > Try with -XX:AggressiveHeap (JDK 1.4) or -XX:+AggressiveOpts (JDK 1.5+) >> > > instead of -Xms and -Xmx. ( >> > > >> > >> http://www.oracle.com/technetwork/java/javase/tech/vmoptions-jsp-140102.html >> > > ) >> > > >> > > I work with a distributed environment with 3 JMeter servers per machine >> > > (CentOS5, 6GB RAM, 2-cores processor, 20GB HD), using 1.5GB heap per >> > JMeter >> > > server. This configuration hold almost 300 threads per JMeter server >> with >> > > no problems. >> > > Sometimes a big heap mess up more than help. >> > > >> > > Regards. >> > > Flávio Cysne >> > > >> > > 2012/7/27 Deepak Goel <[email protected]> >> > > >> > >> Hey >> > >> >> > >> You might need a bigger hardware box, i guess :) >> > >> >> > >> :) >> > >> Deepak >> > >> >> > >> On 7/25/12, Philippe Mouawad <[email protected]> wrote: >> > >> > Hello, >> > >> > Read this : >> > >> > http://jmeter.apache.org/usermanual/best-practices.html >> > >> > >> > >> > Ensure you are not using View Results Tree or any other costly >> > listeners. >> > >> > >> > >> > After checking all these points don't jump to conclusion that other >> > >> > parameters are OK if your original load test tool is reporting OK. >> > >> > Built-in samplers are rather highly tested and used , so i am not >> sure >> > >> > writing your own will help. >> > >> > >> > >> > Regards >> > >> > Philippe M. >> > >> > >> > >> > On Tue, Jul 24, 2012 at 8:28 PM, unjc email <[email protected]> >> > >> wrote: >> > >> > >> > >> >> Hi all, >> > >> >> >> > >> >> I have a test plan with 100 threads of HTTP Request samplers >> hitting >> > a >> > >> >> web server via proxy server. My goal is to hammer the server as >> fast >> > >> >> as possible for an hour. When I run the test for first 30 seconds, >> > >> >> the throughput is excellent; slowly I found it keeps dropping. I >> am >> > >> >> sure it's not a problem with the network, proxy or web server >> because >> > >> >> the other load-testing tool has been working fine before I switch >> to >> > >> >> try Jmeter. >> > >> >> >> > >> >> I found that Jmeter consumes lots of memory and cpu cycles during >> the >> > >> >> test. I have tried "Statistical" and "Stripped" mode, along with >> > >> >> 2048M heap size. Nothing really helps to keep the test running for >> > >> >> long period of time. The java process is one hungry monster in the >> > >> >> top stats. Is there any lighter sampler available (that supports >> > >> >> proxy, custom originating IP, and https)? If not, is there any >> good >> > >> >> document that could help start writing my own plugin? Any comment >> > or >> > >> >> advice will be greatly appreciated. >> > >> >> >> > >> >> >> > >> >> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ >> COMMAND >> > >> >> 6128 root 20 0 10.5g 2.3g 14m S 539.2 19.5 6:16.95 java >> > >> >> -server -XX:+HeapDumpOnOutOfMemoryError -Xms2048m -Xmx2048m >> > >> >> -XX:NewSize=128m -XX:MaxNewSize=128m -XX:MaxTenuringThreshold=2 >> > >> >> -Dsun.rmi.dgc.client.gcInterval=6 >> > >> >> >> > >> >> >> > >> >> >> > >> >> Thanks, >> > >> >> Jacky >> > >> >> >> > >> >> >> --------------------------------------------------------------------- >> > >> >> To unsubscribe, e-mail: [email protected] >> > >> >> For additional commands, e-mail: [email protected] >> > >> >> >> > >> >> >> > >> > >> > >> > >> > >> > -- >> > >> > Cordialement. >> > >> > Philippe Mouawad. >> > >> > >> > >> >> > >> >> > >> -- >> > >> Namaskara~Nalama~Guten Tag~Bonjour >> > >> >> > >> >> > >> -- >> > >> Keigu >> > >> >> > >> Deepak >> > >> +91-9765089593 >> > >> [email protected] >> > >> http://www.simtree.net >> > >> >> > >> Skype: thumsupdeicool >> > >> Google talk: deicool >> > >> Blog: http://loveandfearless.wordpress.com >> > >> Facebook: http://www.facebook.com/deicool >> > >> >> > >> "Contribute to the world, environment and more : >> > >> http://www.gridrepublic.org >> > >> " >> > >> >> > >> --------------------------------------------------------------------- >> > >> 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]
