BraydonT wrote:Can you tell me more about the "duty cycle" speedup approach that you talk about? How does it work and what kind of load multipliers does it get you.
Suppose you have a test script that emulates actual user behavior, and it includes user waits (think times) of 90% of the total playback time. That is, only 10% of the time is the browser doing anything other than waiting.
That script has a duty cycle
The workload imposed by this script in terms of bytes downloaded is some value, say X.
If you take the same script and change it so that the wait times are eliminated completely, and then run the script 10 times in a row, it will take as long as the unmodified script but will download 10X the number of bytes.
From the server's point of view, the first case seems to be one user and the second case seems to be 10 users in the same length of time.
So you can simulate 10X the number of users of 10% duty-cycle scripts just by removing all of the think times (wait times).