jaxsonr wrote:How important is "realism" in load testing? I mean, if I have a sequence of URLs that I can retrieve very rapidly, isn't that the same load on the server as having a lot of users?
It could be, and then again it may not be.
Here are some things "wrong" (really, not wrong, but inaccurate and therefore potentially missleading):
(1) Just downloading a URL by itself is NOT the same as downloading the page and all of its constituent URLs. Of course, if this is a terminal page the comment doesn't apply...but if typical web pages involve 10-50 component files then you may have a very long list to get all of them.
(2) Serial download is a poor approximation of the way browsers download pages, which is done in parallel with multiple HTTP sessions all active at the same time. Two successive downloads of the same web page may involve widely different paths through the web, and usually complete the page component downloads in different order. Your serial download of a set of URLs does not simulate how a browser really sees things.
(3) If your web page involves any AJAX then you may not know in advance what the URL sequence actually is...that's the nature of AJAX. That's why a recording of an AJAX session records past experience, but not present experience. Even worse, the sequence of URLs from the current experience may be quite different each time to perform the action.
All of which are reasons why the realistic approach, using the browser to generate the load that the browser normally generates, is so much more realistic in terms of applying load.
eValid LoadTest Support
Posted: Mon Dec 05, 2011 4:53 pm