I keep hearing I need to run 100s or 1,000s or 10,000s or 100,000s of virtual users to find out what's failing?
Is that really necessary?
Morning SandraC. Thanks for posting.
We don't think so. Not really.
The issue resolves better if you think of it in terms of realism of the tests.
Typical Virtual Users (VUs) use the HTTP/S protocol to impose load on the server. But because the agent is not interacting with the server -- as the browser does, especially in the case of AJAX (and what isn't AJAX these days?).
So, real sessions are maybe 100X the data load compared with VUs. A factor of 100X is far to big to be ignored.
Sure, you get lots of more "virtual users" but you also get lots more "virtual load", which by comparison with the eValid approach of using real browsers.
But the real message is in the results. In practice, running many many tests, what we have found is that you very often see rolloff in performance at as few as 500 users..
So, what's the need for 10,000 VUs if you can get the server to fall over with as few as 500?
-- eValid Support