Github user tdunning commented on the issue:
https://github.com/apache/zookeeper/pull/307
So here is an alternative suggestion.
For each class of operation (GET, PUT, MULTIPUT, DELETE and so on) maintain
a data structure (such as t-digest or Floathistogram see
https://github.com/tdunning/t-digest) that keeps track of the distribution of
latencies for that operation.
Log any transactions that are above a settable percentile latency for
operations of that class. Default could be 99.99%.
Operations longer than 2x the 99.9%-ile latency should not be entered into
the historical distribution.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
