Personally I've used WatirGrid[1] with great success, I used it to drive 50 browsers concurrently. 50 is not the ceiling, I was just limited by how many boxes I had at hand.
The limitation with browser based testing is the resources required on which to drive the browsers! BrowserMob have reduced this (but they only support Selenium). I'm thinking of something similar for Watir users. I've also been experimenting with headless tests (watir-webdriver with htmlunit) with a fair amount of success, which increases your bang-for-box. I'll talk about some of this at Watir day in any case. In my experience, not every load test is thousands of users. I've load tested some awful intranet apps that have fallen over well before 100 users ... So you may get some mileage out of Watir yet ... Protocol level testing is probably better for load tests with thousands of users though based on bang-for-box. I expect there's an undiscovered sweet spot somewhere there where the cost per vuser shifts in favour of protocol level (when using commercial licensed tools) testing. Cheers, Tim @90kts https://github.com/90kts/watirgrid/blob/master/EXAMPLES.rdoc On Tue, Mar 15, 2011 at 4:21 AM, Chuck van der Linden <sqa...@gmail.com>wrote: > Generally speaking, in order to handle a large pool of threads and be > able to scale the load, most loadtesting is done at the protocol > (HTTP) level, not using full blown browser based clients. > > Since most load is not from simple page hits against static or cached > pages, you need to actually be interacting with the server as would a > large number of unique users. Most decent webservers can serve cached > copies of a static page until the pipe saturates without overloading > the server or the back end. So having 200 users doing the same > search, doesn't create the load you need. > > So what you need is a system where thread to use its own unique data, > so you can simulate a hundred different users, shopping for 80-100 > different items, or doing different searches or whatever. (and 100 is > just to start, a true loadtest often simulates interacting with many > thousand users over a period of an hour or so) > > While I might use Watir to conduct a functional test DURING a > loadtest, I would not use it to apply the load. > > Despite the existence of the example, I personally would not do > loadtesting with watir. I'd use something like the loadtesting tool > in VS2010, Jmeter, or if you can afford it, Loadrunner. (which one > would work best and provide the best bang for the buck depends on your > server environment) > > On Mar 11, 10:30 am, "pierrelebai...@gmail.com" > <pierrelebai...@gmail.com> wrote: > > Hi, > > > > to see there are realy diffrent thread that are running change : > > ...("Thread #{i} executed... > > by > > ...("Thread #{Thread.current} executed ... > > > > in log file information for loadtester.rb which can be download here : > > > > http://wiki.openqa.org/display/WTR/Re-Usable+Load+Testing+Example > > > > best regards, > > plebailly > > -- > Before posting, please read http://watir.com/support. In short: search > before you ask, be nice. > > watir-general@googlegroups.com > http://groups.google.com/group/watir-general > watir-general+unsubscr...@googlegroups.com > -- Before posting, please read http://watir.com/support. In short: search before you ask, be nice. watir-general@googlegroups.com http://groups.google.com/group/watir-general watir-general+unsubscr...@googlegroups.com
