This procedure is used when customers’ ports on the MIX switches are virtually cut at lower speeds (virtual rate-limit-VRL); MIX runs monitoring tools on each router interface connected on the peering LAN.
Every month, the data generated by the monitoring system are processed through an algorithm that cuts off the first 10% of the peak values and:
a. if the bandwidth usage is less or equal to the VRL value, no action is made by the MIX;
b. if the bandwidth usage is greater than the VRL value:
- after the first month, MIX sends a warning message by e-mail to the technical contacts;
- after the second month, enduring the over-usage, MIX requires a balance and adapts the VRL value to the closest upper one (or to the port speed if there aren’t intermediate VRL values);
MIX sends to the ISP administrative contacts an invoice for the difference between the fee already settled and the fee related to the new value, computed from the first month of over-usage;
- if during the second month the bandwidth usage drops below the VRL value, the process is cleared;
- if at the end of the third month the ISP has not paid the adjustment invoice, MIX configures on the switch port a physical rate-limit set at the VRL value, noticing it to the technical and administrative contacts of the ISP company.
Note: the third month has to be intended as a deadline for meeting payments: even if during the third month the bandwidth usage drops below the VRL, the VRL of the ISP is administratively update to the new calculated value.
Description of the monitoring algorithm
The considered values are those produced by the monitoring tool used by the MIX for checking the connection circuits state and usage on its own LAN. The monitoring tool is, today one of the most popular public domain software for this purpose.
- MRTG gets monitored interfaces counters values and stores them into log files, particularly the values of the couple (Max IN, Max OUT).
- Every 5 minutes the maximum values of the counters, extracted every minute inside the range, are calculated; these values are stored in a log file (one for each interface) closed every 24 hours: in this way it’s possible to make a daily analysis of the bandwidth usage with a 5 minutes granularity.
- The algorithm extracts the couples (Max IN, Max OUT) from the daily log file and sorts them in two decreasing arrays; the first 10% is cut off from each array and the remaining higher value is stored in the monthly historical file related to that organization.
- At 00:01 of the 1st day of the following month, the historical file is closed and sorted in two decreasing arrays (IN and OUT); again, the first 10% is cut off from each array and the maximum value between the remaining maximum inputs and outputs is extracted: this value is the value considered as the bandwidth usage for that month.