Measuring bandwidth using the 95th percentile is a common method used in network monitoring and billing to determine the amount of bandwidth consumed by a network connection or server over a given period. Here’s an explanation of how it works:

1. Bandwidth Measurement: Bandwidth refers to the amount of data that can be transmitted over a network connection in a given time. Measuring bandwidth involves monitoring the amount of data transferred over a specific connection or server.
2. Sample Period: The bandwidth measurement is typically taken over a defined sample period, such as one month. During this period, the data usage is continuously monitored.
3. Data Collection: Throughout the sample period, the amount of data transferred over the connection is recorded at regular intervals, usually in intervals of 5 minutes.
4. Sorting Data: Once the sample period is complete, the data measurements are sorted in ascending order, from the lowest usage to the highest.
5. Discarding Outliers: In the next step, the top 5% of the measurements, also known as outliers, are discarded. This step helps remove any unusually high spikes or short-duration bursts of data usage that may skew the overall measurement.
6. Calculation: The 95th percentile is then calculated based on the remaining data measurements. This percentile represents the data transfer rate that is exceeded or equalled for only 5% of the sample period. In other words, it indicates the sustained level of bandwidth that can be provisioned without the need for continuous overage.
7. Billing or Monitoring: The 95th percentile measurement is commonly used for billing purposes in internet service provider (ISP) or hosting provider scenarios. It allows service providers to bill customers based on their actual bandwidth usage, while avoiding penalizing them for occasional traffic spikes. It is also used in network monitoring to assess network performance and identify potential congestion or capacity issues.

By using the 95th percentile measurement, bandwidth billing and monitoring can be more reflective of the customer’s actual usage patterns, as it focuses on sustained usage rather than short-term spikes. It provides a fair and representative measure of the bandwidth consumed, allowing for more accurate billing and better capacity planning.