[httperf] request rate steps

[Rick Jones]

If you are concerned about the load going to zero for short periods of time, if 
you are going to be scripting things anway, might script it to add additional 
instances of httperf rather than killing the old one and starting a new one at a 
higher level.  You need to make certain the earlier instances are configured to 
run "long enough" or perhaps "forever" and rely on a signal to tell them when to 
cease and/or emit interim results.

How easily that is done with httperf I've no idea, but if it can be done that 
would address your concerns about the load tapering off between levels.

Specifically what are you concerned about anyway?  A great many benchmarks work 
the way previously described - stop one load leve start another higher one - so 
if there is indeed a non-trivial problem it might be good to discuss it further.

rick jones