Why do you insist to run your load tests without using the cache ... doesn't that mess up the accuracy of the tests?
I'm concerned about a lot of false data during load tests.
Good question BJustman.
Indeed "cache management" is a critical issue in server loading experiments.
In designing the eValid load testing solution we chose a conservative approach:
Better to over-load and err on the side of having a bit MORE capacity than to under-load and have a false sense of security.
That principle works out to require that the tests that run in parallel and on multiple machines NOT take advantage of already-loaded pages.
Normally the browser will "remember" a page that was previously loaded and not bother to re-request it from the server (unless the age of the copy on the server is younger than the age of the one in the cache).
eValid has specific cache management capabilities, described here: http://www.e-valid.com/Products/Documen ... ement.html
In eValid, to avoid any bias in the data, we specifically have LoadTest runs set up to NEVER use the cache.
This is an option that is available on a single-user (non-LoadTest) playback, as you can see in the documentation.
But we further automatically set that internal switch in every *evl (LoadTest) run, even those that are launched from the command line.
It is true that this means that a page that is visited twice in a playback would be loaded twice, rather than just once -- as it would be in normal operation (the second load would be from the cache).
That's more work, and thus this is a over-load of the server.
We think it is a good trade off.
-- eValid Support