We have a strange case between Utilization vs. Utilization-In, and Utilization-Out. It seems that the Utilization of the router interface is the same as Utilization-In metrics, and not the combination of Utilization-In and Utilization-Out. This is for both the Average and Max average.
Here is some info on how we do the calculations:
max=avg when looking at the polled value read from device.
As data comes in, we set max/min/ave to same value. when we rollup into hourly, we use the rollup strategy in the metric family to either SUM/AVG the as-polled data into _ltd table for the hour.
So, we take the min and max for the hour and store that as 1 entry for the hour. For daily, we find the min and max and rolled up avg/sum for the entire day from the as-polled data and store 1 entry for the day.
Weekly table rolls up the daily values, so 7 data pts into 1.
So, takes the min/max of those 7 data points, and either ave/sum of the 7 data points (so avg of an avg).
you would have to be polling the same values every time for them to be the same for hourly/daily rollups.
At this point, the next steps are to enable detailed poll logging for a problem device and see what the raw data coming in looks like.
If we are indeed getting a static value, than that would explain this.