SmartBear WeloadUI doesn't stop

Question

I'm new at WebloadUI and although seems to be an amazing tool, I'm having problems simulating concurrent users.

I record the test using the methodology explained in the SmartBear documentation, I verify the scenario to check every requests has its respective correct reply. Everything works just fine.

The problem comes when I try to add virtual users to the mix. I set the test for 10 concurrent virtual users, and then push start. The graphics show how the amount of virtual users ramps up to 10 (as expected) and then, that number lows to 7 or 6, then to 4 or 3, and stays there for the next 15 minutes or so. When the test finishes the result shows a lot of timeout responses, likely because the stuck threads.

Of course I have assumed that the server was overloaded, for that reason the timeouts. But that seems very unlikely since the test was recorded against a Wikipedia article and, as I said, I used 10 virtual users. I really don't believe Wikipedia cannot handle 10 virtual users against one page.

What could be happening? Thanks in advance.

Answer 1

The problem might be with SmartBear, or with Wikipedia, or with your network connection. You did not say how frequent those virtual users were hitting that page. It is unlikely that Wikipedia is incapable of handling more than ten concurrent sessions against that same page. However, if you are hitting the page fast enough, it is conceivable that Wikipedia is treating those hits as a network attack, and is intentionally throttling its responses to you. Here is one way to find out: while the test is running and slowing down, open a browser on the same machine (or at least on the same IP address) and go to that page. If your browser is slow, SmartBear is not the problem. If the browser loads the page quickly, SmartBear is probably at fault.

Answer 2

To add to @user246's excellent suggestions - I'd suggest trying against a test site running on your system or in your local network, preferably one you have full control over.

With Wikipedia, there are too many other variables to deal with: the performance of the Internet at large, how Wikipedia's servers handle a bunch of requests coming from the same machine (they probably assume a splat of requests from the same IP address is a denial of service attack), the performance of your local network gateway and so forth.

If you can eliminate some of those unknowns, you'll probably have more success with your testing (caveat here: if you're doing this with your employer's network, keep close tabs on what's happening. I know someone who brought down the company LAN when he ran a load test against the company's web store application. Of course, he was running it with 300 virtual users...)