This will give the java process some help in memory management as it will have 128m of heap instead of the default 64m.
Regards, Vishal.
peter lin wrote:
I'm not sure it's JMeter's fault. I've compared apache bench to JMeter. When I use AB to load test a jsp on a 2Ghz AMD box with 1gig of ram, anything more than 75 users yield unreliable results. AB definitely uses less memory, but over 50 users on a system with just one ethernet is going to be limited by network I/O. It's usually better to load up a couple of systems with 25-50 users.
I should qualify what I mean by unreliable. Basically the variance/deviation between requests was considerably higher. In many case I had to use AB with keep alive to get reliable results. Without using keep alive, the total time between each run would vary dramatically. If I ran a large number of tests and averaged the results, it yielded better results, but takes a lot more time.
peter
Duncan Frostick <[EMAIL PROTECTED]> wrote:
That's Java for you. As great as the portability and libs are, it's such a painfully slow lang for performance tasks. I have a Solaris server with 2GB of RAM and it can't do 100 threads without giving me stupidly high response times (of close to an hour...) on some requests.
Its a pity because JMeter is such a fantastically good tool when it works, its just let down by performance issues. If you want to load test at anything like a high load with JMeter you need several high powered servers doing it in unision to spread the threads.
I was trying to use the GCJ compiler - which claims to compile java bytecode to machine code - to increase performance but I couldn't get it working, that could be a possible performance boost. My general rule of thumb is that for anything above 75 users JMeter/Java is too resource hungry to give accurate results on my hardware.
And now I'm trying high-loads, thats a problem. However, IBM have a nifty if hard to configure tool called 'stress'. It's written in C so it performs very well, just lacks a lot of JMeters great functionality. You can find it online if you serach for IBM Web Performance Tools.
Perhaps a port of JMeter to C++ should be considered? I know this would involve doing a lot of painfully boring networking and threading code but the performance gain would be huge.
Cheers, Duncan
Mark McWhinney wrote:
Hi all,
I am a long time LoadRunner user but am new to JMeter. I am impressed with its functionality. It is a lot closer to LoadRunner than many of us in the LoadRunner community realize.
I am going through the usual newbie issues but am actually reading the manual and figuring it out.
I ran a test with one thread group with 30 threads that had six HTTP requests with a 5 second wait time between each request. I maxed out the CPU at about 20 to 25 users on a 800 MHz Windows 2000 machine. Memory utilization was about 350 Mbytes. Is that typical? I am really going to need 10 machines to do a 250 user load test or am I missing something?
--------------------------------------------------------------------- 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]
--------------------------------- Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
