6500 CatOs command sh port mac util and MRTG/SNMP values

Discussion in 'Cisco' started by H.U.A. Koers, May 28, 2004.

  1. H.U.A. Koers

    H.U.A. Koers Guest

    Hi, i'm confused about the 5 minutes average values from the 6500 CatOs
    command SHOW PORT MAC UTILIZATION (times 8 because of octets) and the values
    gathered by MRTG/SNMP with default config (cfgmaker and bits option).

    Does anyone know how the values, shown with sh port mac util, are
    calculated? They do not seem to match the values calculated by MRTG/SNMP.

    When I calculate the average over aprox. 5 minutes (Transmit2-Transmit1)*8
    (octets) /300 (seconds), with the values from the command sh port mac, it
    matches MRTG again. So MRTG seems to calculate the right values...

    Search on cisco.com on the catos command show port mac utilization does not
    give any answers.

    TIA, Arjen
     
    H.U.A. Koers, May 28, 2004
    #1
    1. Advertisements

  2. H.U.A. Koers

    Tax Johnson Guest

    Does the 6500 show more than 100 Mb/s being used on the port? If so, you'll
    need to tell MRTG to use 64-bit counters. You do this by telling MRTG to
    use SNMPv2c.

    http://people.ee.ethz.ch/~oetiker/webtools/mrtg/mrtg-reference.html
     
    Tax Johnson, May 29, 2004
    #2
    1. Advertisements

  3. H.U.A. Koers

    H.U.A. Koers Guest

    Correct! Normal counters do reset too often to get satisfied results, when I
    use the HC / SNMPv2 option :)::::2 behind every interface) results become
    more visible / clear. Can use the snmp version option in cfgmaker to tell
    MRTG to always use SNMPv2

    Thanks Tax!
     
    H.U.A. Koers, Jun 1, 2004
    #3
  4. H.U.A. Koers

    AnyBody43 Guest

    I agree with the counter overflow issue described earlier,
    however there is another issue.

    The 5 min (or whatever you set the "load interval" to value is
    NOT a 5 min (or "load I..) average.

    It is IIRC an Exponential Weighted average with a Time Constant
    of 5 (or whatever) minutes.

    In my experience with the default 5 min "load interval", if the
    actual load is changed from say 0% to say 50% suddenly,
    the displayed load changes for between 15 and 20 minutes.

    The exact behaviour is described in:
    http://www.cisco.com/warp/public/66/3.html
     
    AnyBody43, Jun 1, 2004
    #4
  5. H.U.A. Koers

    H.U.A. Koers Guest

    Thanks for that info, IMO the calculated "average" has indeed sort of an
    exponential factor. When a link suddenly dies, the average smoothly fades
    away instead of just becoming 0. The same is with spikes in the network.

    Arjen
     
    H.U.A. Koers, Jun 2, 2004
    #5
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.