Merge lp:~lucio.torre/txstatsd/use-uniform-sampling into lp:txstatsd
Proposed by
Lucio Torre
Status: | Merged |
---|---|
Approved by: | Sidnei da Silva |
Approved revision: | 91 |
Merged at revision: | 91 |
Proposed branch: | lp:~lucio.torre/txstatsd/use-uniform-sampling |
Merge into: | lp:txstatsd |
Diff against target: |
278 lines (+71/-79) 4 files modified
txstatsd/metrics/timermetric.py (+22/-29) txstatsd/server/configurableprocessor.py (+2/-1) txstatsd/tests/metrics/test_timermetric.py (+6/-22) txstatsd/tests/test_configurableprocessor.py (+41/-27) |
To merge this branch: | bzr merge lp:~lucio.torre/txstatsd/use-uniform-sampling |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Sidnei da Silva | Approve | ||
Review via email: mp+106880@code.launchpad.net |
Commit message
Drop exponential sampling for uniform sampling.
Description of the change
drop exponential sampling for uniform sampling.
Between each report() call we create a bucket where we sample uniformly from the values seen in the period.
This will produce more spiky metrics but with more information and an more exact rate, but also means dropping Xmin_rate.
also, cheaper on the cpu.
To post a comment you must log in.
Looks good, works great. Not much improvement in CPU, but meh. +1!