Looking at the timeseries, I see there is MAX ,SUM ,MIN , COUNT, AVG and Perc (1-99). Does this mean that there are 104 value's kept for every minute, 5 minutes, hours and days per measure?
And when you look at the AVG over (for example) 29 days these will be the avg (per 5 min), witch are the avg (from the 1 min)?
If the metric represents a statistical counter, every minute contains MAX, SUM, MIN and MAX. In case many measurements are collected for the metric within a minute (e.g.: service call response times, which can be ten thousands of individual measurements per minute) those metrics also offer a percentile aggregation type. A percentile type of metric stores the percentile function by using a approximation function that can deliver the percentiles from 1 to 99. Its not storing 100 individual percentile values per minute but rather a parameterized function.