I have a topology below (two switches S1 and S2), each interface has two metrics (ifinutilization and ifoututilization).


%ifinutilization = ( (ifinoctets * 8) * 100 ) / ifspeed
%ifoututilization = ( (ifoutoctets * 8) * 100 ) / ifspeed

i want to summarize the link utilization, can someone help me ?

  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer. – Ron Maupin Dec 17 '20 at 15:08

When you summarize periods of utilization, there are two basic options:

Calculate the average: sum all values and divide by their number. Beware - averages don't properly represent uneven load and may hide bursts with congestion behind comparatively low average utilization. However, averaging gives a good estimation of overall data volume. Note that with a low frequency of ifin/outoctets sampling (less than perhaps one sample per minute), you're already averaging.

Find the maximum: find the highest value and drop all others. In contrast, this shows peak loads very well but allows no estimation of actual data volume or overall actual load. In order to see the peaks you need a high sampling frequency (reasonably more than one sample per minute).

  • The thing is we are dealing with a packet switched network here, so at any given instant the line is either 100% used or 0% used. So you have to average over some period, the question is what period. Most monitoring systems I have seen average over at least 5 minutes, which makes it easy to miss fast spikes. – Peter Green Jan 19 '20 at 1:14
  • @PeterGreen Agreed - you wouldn't sample in the sub-second range though so that a single packet causes 100% utilization. I have refined the question somewhat to include time frames. – Zac67 Jan 19 '20 at 13:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.