Hi guys, I've been putting more thought into load testing and what that means for Watir. Watirgrid allows for distributed testing across a grid network, but it is heavy in the sense of 1 core = 1 browser. So a 1000 user load test means a lot of cores spun up on EC2 or the like ...
WebDriver seems the right direction for me. I've since added webdriver support to watirgrid, and am really interested in capabilities of htmlunit. I can easily spin up say 100 threads per core using htmlunit with no obvious degradation to performance. It's not the perfect world (especially with Rhino) but it's a good start. I know others have thought about this and experimented a little. One of the traps I see in this approach is htmlunit itself, or rather Rhino. If the application under test has some weird JS combo not executable by Rhino then the test case is stuck. It would be good if we could gracefully downgrade to a lower (protocol) level if needed. For example if the test case got stuck executing within htmlunit, we downgrade to httpunit or net/http and make the request needed to get past manually. Has anyone experimented with this before? I was thinking of perhaps hijacking the session (dumping cookies from Watir, re-using them with net/http if possible) Jari also suggested extending HtmlUnitDriver but I'd most likely need to be doing this in Java (maybe via JRuby) to get this working ... Thoughts? I'm really keen to get some movement in this space (load tests with Watir, one test API for functional _or_ performance automation etc) and would love to talk about this with anyone else interested at Watir Day (SeConf). Other areas of interest are tying Watir in with other performance monitoring tools (like splunk, dynatrace, fiddler, webtiming etc). I think I'll pursue with Watirgrid until it makes sense to split out otherwise. Cheers, Tim @90kts || watirgrid.com
_______________________________________________ Wtr-development mailing list Wtr-development@rubyforge.org http://rubyforge.org/mailman/listinfo/wtr-development
