In Yammer metrics, the 15MinuteRate provided by the Meter is not a pure average for the last 15 minutes. It's a Exponential Weighted Moving Average. The value is calculated by taking 36% of weighted average from the broker start time and 64 % of weighted average from the last 15 minutes.
I initially thought this metric is the total volume of bytes that came in for that topic in last 15 mins. So, if I collect this metric every 15 mins, then I will have total volume over time by calculating the sum of data points in the time interval. You can use `count` instead of `15MinuteRate` to find out the total volume of bytes that came in a topic for last 15 mins by taking diff of the two points. On Sat, Feb 2, 2019 at 8:39 PM Amitav Mohanty <amitavmohant...@gmail.com> wrote: > Hi > > I am trying to measure incoming bytes over time. I am trying collect the > following metric and apply integral function over a set of data points on a > time series. > > kafka.server:type=BrokerTopicMetrics,name=BytesInPerSec,topic=my_topic > FifteenMinuteRate > > It seems that the number I am getting is much less than expected. So, I am > suspecting that my understanding of this metric is wrong. > > I initially thought this metric is the total volume of bytes that came in > for that topic in last 15 mins. So, if I collect this metric every 15 mins, > then I will have total volume over time by calculating the sum of data > points in the time interval. > > Please confirm if my understanding is correct. > > Regards, > Amitav >
