nickva commented on issue #4650:
URL: https://github.com/apache/couchdb/issues/4650#issuecomment-1603005683
The basic idea for preserving the windowed updates with 10 second windows,
and 1 second granularity, like we have currently, could be to use at 10 counter
arrays (in a tuple for instance, to allow quick iteration), one per second:
```
Counters = {counters:new(NumberOfBins), counters:new(NumberOfBins), ....}
```
So `update(#hist{}, Val)` would do something like `TimeIndex =
erlang:monotonic_time(second) rem tuple_size(Counters) + 1`, `CounterRef =
element(TimeIndex, Counters)`.
In order to for the trimmer/cleaner process to not step on the toes of
current updaters, we'd create 15 or 20 counters, to allow plenty of time in
between when the trimmer process resets the old counters and some delayed
updaters still updating it. Using monotonic time prevents the updaters from
going backwards but there could be some time between them reading the monotonic
time and the update itself.
In this way we avoid the need to modify (delete old and create new) counter
arrays in a persistent term. The persistent histogram term structure remains
immutable. The only thing that gets updated are the counter values themselves
which should be efficient.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
