I'm new to Load / performance testing.
Could u let me know :
1. How is a site / client-server application tested for the Load or the performance issue manually and using automated tool (say, Load Runner) ?
2. Suppose the tester is testing it manually, then how will he be able to know the response time from the server ?
3. While testing a web site, the tester needs to test on various modes, like Dial up connection, Leased line internet service etc., just to check how the system will behave. How will the user be able to decide the accurate response time / evaluate the performance inorder to decide whether the system is behaving as expected ?
I have never performed performance/load testing manually, and outside of testing some sort of batch processing application where I could push various amounts and sizes of batch files through, I don't know how manually testing would be done. That said...
1. Automated tools have the ability to wrap a timer around transactions that will track the time it takes from making a request to a server and the response comes back. They also have the ability to simulate multiple users (called virtual users).
2. Outside of having some software on the client testing machine that would do the time tracking or having scripts on the server that would track the time it takes from when a request is received to the time that the response is sent, the only way I can think of is having a stopwatch and being really quick with your thumb.
3. Two things, some of the automated load testing tools have speed settings that you can set to test at different connection speeds. You could also look into using a WAN emulator. These will slow down the arrival rate of packets coming through it to simulate different connection speeds. The second part about how to decide whether the performance you are seeing on the lower speeds is good would be a business decision. The business could either set the requirements ahead of time so that you know whether or not the performance is ok, or you could run the tests and take the results to the business/project owners to let them decide if the response times are reasonable.
We once did a manual loadtest on a web application. We used 20 people following written out scenarios.
Interestingly this was done after we ran the tests with our loadtest tool. The development team could not believe that the tool did actually simulate 20 users in a correct fashion and the doubted our observations on the deadlock problem the application exhibited. Anyway, as it turned out that problem popped up during manual loadtest as well and really left the whole group with quite a bad impression of the quality of the system under test. Needless to say that we have not been challenged on our loadtest results and analysis again.
[This message has been edited by rstens (edited 12-06-2002).]