gBennett wrote:I really like your approach to creating server load, but don't you guys get accused of doing a "denial of service attack" for very high loads?
Yes, that has been suggested but the numbers don't work in that direction.
Every eValid playback is a simulated REAL user, generating what, to the server at least, cannot be distinguished from a real user's inputs.
In typical experiments about 1,000 eValid browser users (BUs) are run in parallel on one machine to impose load on a server. To get 10,000 users you would need to use 10 machines (we have done that level of driver).
Such experiments are done under very carefully controlled situations, and uaually on a "test environment." [We are very resistant to driving the real application, for obvious reasons.] And in every case there is someone watching the performance monitors on the server side to try to identify when a particular component or system element hits a saturation point. It's often the case that we turn off such an experiment when some measured quantity exceeds a threshold. After all, the purpose of such an experiment is insight!
On the other hand, a denial of service attack is (a) malicious and (b) intended to saturate servers with any kind of request. We're not malicious, and eValid's requests to the server are generated from scripts and within the
individual BUs that are runnig the script. These are two different animals!
eValid LoadTest Team