tclwebtest theoretically supports running several test runs repeatedly and measuring the time used in total (roughly, in seconds only), optionally in parallel by forking off child processes. The forking appears to be broken though, for some reason the tclx forking calls don't behave anymore like they did initially.
Disadvantages: not highly accurate, quite resource intensive on the client machine.
Advantages: can measure arbitrarily complicated user actions, not only simple page requests.
