jmeter latency vs actual browser load test

6
votes

Is this a valid testing for checking of how much time to load a web under test with 500 concurrent user.

I run jmeter with a 500 thread user , ramp-up period = 50 and loop count forever. with a listener with "results in table" that also record the latency.

While jmeter is running, i try to load/browse the web under test using actual browser(in my case IE8) , and it loads in 7 secs. but based on the latency the majority of result is 50k++.

is the 7 secs load time in actual browser is consider a "response time result"? since it is load in actual browser.

another question: is the latency 50k is converted to sec? means 50secs. to load the web under test if we based on the jmeter result? kindly clarify this to me please :)

1
performancebrowserloadjmeterperformance-testing
With 500 users you should use the command line (-n) mode of jmeter instead of GUI, and remove the results in table listener for better performance and more accurate results. Use the "-l" parameter to log the results in a JTL file. (you can later load these files in any listener you want, as long as you save the needed fields). You can define which fields should be saved in the result file, with the options in jmeter.properties file, lines starting with "saveservice.". You might want to consider saving XML so you can actually see the requests/responses received. - Dominik Antal

1 Answers

11
votes

In simple words, Latency is network delay (time taken by network while transferring data)

In JMeter latency is time between, when request is sent to server till first byte of response reaches the client/Jmeter. If response time is very low enough then you wont get precise measure of latency. If Response time is high then probably you will get correct measure.

In Jmeter Latency shares the measure as response time i.e. ms/seconds.

Your 7sec in browser is (Response time (Processing time + Latency) + Rendering time). In Jmeter rendering time is not present (As it is not a browser). Though your rendering is very low as compared to response time but in cases heavy content websites rendering time is comparable. Thus should be considered.

I hope this clears your doubts :)