Open ruiaraujo opened 5 years ago
@ruiaraujo honestly skipper should not reinvent the wheel but use Prometheus, which I think already does this.
DropWizard style metrics work good for a single instance, but not for N > 1, because you have to aggregate the exposed metrics using max()
.
Prometheus is a completely different model, DropWizard style metrics are in the codebase and we should have the best quality possible of this type. 😄
Is your feature request related to a problem? Please describe. Skipper currently has either EDS or Uniform reservoirs which suffer from various issues specially in the face of low traffic.
https://medium.com/hotels-com-technology/your-latency-metrics-could-be-misleading-you-how-hdrhistogram-can-help-9d545b598374
Describe the solution you would like Skipper should include the option for using HDR based histograms which should provide more accurate values.
Describe alternatives you've considered (optional) Tracing data
Additional context (optional) Add any other context or screenshots about the feature request here.
Would you like to work on it? Yes, but no time