[mrtg-developers] Measurement Mismatch between MRTG reports and PacketShaper reports.

Anuj abahl at hughes-ecomm.com
Thu Dec 27 12:55:02 MET 2001


Problem Description: 
I have MRTG running in my network and it is measuring the traffic on WAN interfaces of the cisco 7200 series router.
Just after the Ethernet port of the router we have deployed PacketShaper bandwidth management/shaping device.
I can get the bandwidth utilization reports form PacketShaper as well as MRTG.
The main problem is that the Peak Utilization reports from both are not matching.
I have got the following document from the PacketShaper site regarding the measurement technique:

In general terms, Average Rate is the average bits per second during the interval and Peak Rate is the Peak bits per second during that interval.

For all Packetwise enabled devices, for both current and peak rate calculations, we use a proprietary algorithm (EWMA) - exponential weighted moving average. The interval for the EWMA is approximately one second.

For average rate, we use EWMA to get the average rate for that second. We take x samples during that second and use EWMA to get the average for that second. We calculate the average rate for 1 minute interval using EWMA on all one second data. For intervals longer than 1 minute, the Average Rate is calculated by applying the EWMA formula to the 1 minute averages. So the average rate always remains the average during the interval.

The peak rate is also calculated using an EWMA, we take x samples of the peak values during that second and use EWMA to calculate the peak for that second. However, we do not use EWMA algorithm any further. The peak rate for 1 minute period is the highest value during that period (it could have occurred in 25th second or 59th second). For the interval longer than 1 minute, it records the highest value during that interval (e.g. for 1 hour period, it is the Peak among 3600 different values). So the Peak Rate will always be the Peak within that interval.

Could anybody help me in dignosing as to what is causing the mismatch in the reports generated from these two different ways.

Anuj Bahl

Unsubscribe mailto:mrtg-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg-developers

More information about the mrtg-developers mailing list