The first one is similar to Unix's load average and would be very useful in environments with tens of thousands Sidekiq executions per day.
In these scenarios, the single "all-day average" would not be changed in the short-term if the workers take longer to do their jobs.
And the percentiles metrics (maybe p50, p75, p90 and p99?) would be useful in scenarios where the standard deviation of worker's executions is very large.
Hi everyone!
First of all, thanks and congratulations for sidekiq-statistic! It's awesome and very useful!
I'm opening this PR to suggest some new metrics: Exponential moving average (EWMA) and percentiles of the worker's processing times.
The first one is similar to Unix's load average and would be very useful in environments with tens of thousands Sidekiq executions per day.
In these scenarios, the single "all-day average" would not be changed in the short-term if the workers take longer to do their jobs.
And the percentiles metrics (maybe p50, p75, p90 and p99?) would be useful in scenarios where the standard deviation of worker's executions is very large.
What do you think about it?