95th percentile is a well known statistical standard to discard maximum spikes of poll data points which won’'t consider top 5% of data points. In other terms it discards poll data points which are greater than 95% of numbers in given poll data points set.
Example:
Consider CPU utilization data poll points are : 20, 18, 19, 24, 30, 25, 22, 18, 23, 19, 20, 21, 26, 29, 30, 27, 40,35,22, 20 - total 20 data points
Now 95th percentile is number of points which are lesser than top 5% (N) : 95 x 20 /100 = 19
Let us re-order the set:
18, 18, 19, 19, 20, 20, 20, 21, 22, 22, 23, 24, 25, 26, 27, 29, 30, 30, 35, 40
95th percentile average considering 19 points: 23.57 and the unit depends on monitoring parameter.
There is a setting key under global settings with name ’ BSM Report Nth Percentile '. This is by default set to 95 to show 95th percentile. This setting can be changed as per user convenience to show Nth percentile value (N denotes 95 by default and customers can change it to 50 or 75 or whatever values as per their need).