denar90
commented
5 years ago Sorry for the delay. Sounds good, but it requires enhancement for the config we already have and some rework on median calculation. We are not breaking anything here so I'm more than ok with it.
Are you interested in adding this one via PR?
Given the variability in results from a n-samples run, would it be possible to specify a percentile sample instead of median (since intrinsically median can be skewed by an abundance of high / low sample counts).
Use case would be: 1) User defines key metric for percentile analysis and desired percentile (Ex: 80% TTI) 2) Run N samples (preferably 10 or more) 3) Program determines the specified percentile metric and directs user to the sample closest to the derived value.