[thelist] Stress Testing?

[Christopher Marsh]

Casey

> I've been asked by my boss to stress test a new web app they plan to 
> launch soon - they want to know how many users it will handle before
> performance starts to degrade.  Never done this before, and hoping
> someone can give me some pointers.

[..]

> Have any of you ever done something like this before?

I used to use a tool called WebLoad (http://www.webload.org/), which was easy to configure and use. There used to be a free community version which was unsupported and had limitations. I think they *may* have discontinued this now though, and the professional version costs. I now use JMeter, which I've found very useful, although I did have to figure out how to change some memory settings in a configuration file to get it to work.

> As the app is not yet available outside our network, the testing must be 
> done from an internal machine (or machines). So no external cloud
> options will work.

In order for the stress testing to be a useful exercise, you need to be completely clear on what exactly you're testing. It's perfectly possible that the application performs exceptionally within your network, but as soon as it's deployed to the production environment it slows to a crawl or breaks under minimal load. Likewise, it's possible that it performs badly within your network (but not in production) causing you to spend time fixing issues that are actually environment specific. One activity that *will* be useful is to load the application until it breaks, then see where it broke. This will identify any bottlenecks, and you can make an effort to ensure that these are mitigated.

HTH

-- 
Regards

Chris Marsh