[netperf-talk] should service demand replace CPU util in confidence calcs?

[Rick Jones]

Folks -

While I sit here waiting for another test on a competitor's system to not hit 
the confidence intervals on throughput, and local CPU utilization, I am left 
wondering if it wouldn't be better to use the results of the service demand 
calculation, rather than the CPU utilization, when checking confidence intervals?

CPU util will jump up and down whenever Throughput jumps up and down, but in 
theory, the service demand should be a bit better behaved, being the 
normalization of throughput and CPU util.

Or is it bad to use a derived value there?

thoughts?

rick jones