On 26/09/2007, Christiaan Lamprecht <[EMAIL PROTECTED]> wrote: > Thanks, I'll have a look > > You're right the mode=Hold is too memory intensive but I'm not sure I > get the right results with the Statistical mode though: > > > Test Plan > Thread Group (users 2, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 4, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 6, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 8, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 10, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 12, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 14, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Thread Group (users 16, ramp 10, count 1000) > HTTP Request HTTPClient (SSL request) > Simple Data Writer > Uniform Random Timer (Const Delay offset 200, deviation 10) > > > ----------- Elapsed time ----------------- > 2 Mean: 11967 > 4 Mean: 6880 > 6 Mean: 5215 > 8 Mean: 5715 > 10 Mean: 7807 > 12 Mean: 8705 > 14 Mean: 10663 > > > > It should go from low to high right!? And 11 seconds is quite a lot.
One would expect the elapsed time to increase. > Does JMeter get much more efficient with more threads? Not as far as I know, though any overheads will obviously have more effect for shorter tests. But I would have thought 1000 loops is big enough to swamp the overheads. Does the server improve? I can't explain the results. There are a lot of variables here; all I can suggest is trying to eliminate them. For example, what happens if you use plain HTTP instead of SSL? Also try using multiple independent non-GUI runs. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
