As Mike says, batch (non-gui) mode is much more efficient than remote. The network connection between the controlling machine and the server can also prove to be a bottleneck.
== To help test whether your server(s) have enough raw power to drive the test, you can replace the HTTP samplers with a JavaTest sampler - this creates samples without needing the network. You can simulate response times as well as response data, so you can also test the effect of response assertions and log file I/O. If the server has problems running with JavaTest samplers at the desired rate, it won't be able to cope with the additional load imposed by HTTP samplers. Unfortunately, I don't have any figures for the extra load imposed by the more complicated HTTP Samplers (anyway this is likely to vary quite a bit between applications), but it should give you some idea. The next bottleneck is likely to be the network ... S. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Sent: 30 October 2003 15:21 To: JMeter Users List Subject: Re: high-load remote testing The controlling machine is very much a bottleneck, and the remote testing needs some attention before it will handle the kinds of loads you want. Until then, running the separate instances separately will give you your best results. -Mike On 29 Oct 2003 at 15:54, Veronica Baiceanu wrote: > Hi, > > Our company will stress-test a commercial web application, and we are targeting a rate of 1700 hits/second on the application server. We are currently estimating whether we can use JMeter to achieve our goal. So far, I have tried JMeter remote testing using 2 remote JMeter servers, and one controller. I haven't encountered major problems with this setup. However, it would help me to know if anybody successfully used JMeter for achieving a throughput comparable to the one we have in mind. > > > > It would be particularly helpful to me to get some insights regarding the following issues: > > > > - I read that the controller can become a bottleneck in remote testing. Several suggestions were made to solve this problem. The ultimate solution would be giving up remote testing altogether. Instead, each test would be launched by hand from each machine, the times should be coordinated, the results gathered in CSV files, and then merged to get the overall results. I believe I might need at least 5 JMeter servers to get my workload, possibly even 10. I would rather use the remote setup. I wonder if anybody managed to run JMeter in the remote setup, with so many machines, and with my targeted workload. > > > > - It would be helpful if I could get any recommendations regarding the hardware I should use for the testing, i.e. the minimum indicated CPU speed and memory capacity for the machines. We could get 5-10 decent machines that we could dedicate to testing. I wonder if the JMeter resource consumption for large-scale tests could become so critical, that JMeter simply cannot be used. I would not expect the CPU consumption to be higher than for test tools written in C, but the memory consumption should clearly be higher. Any information about particular setups you used would be greatly appreciated. > > > > Thanks a lot, > > Veronica > > > -- Michael Stover [EMAIL PROTECTED] Yahoo IM: mstover_ya ICQ: 152975688 AIM: mstover777 --------------------------------------------------------------------- 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]
