Test Equipment for BER measurement in TCP/IP?

Discussion in 'Cisco' started by Trs80, Jan 28, 2009.

  1. Trs80

    Trs80 Guest

    I have a 100BaseT Ethernet Router to Ethernet Router link over several miles
    of fiber using TCP/IP with fixed IP addresses. Nothing in between except
    fiber and electrical to optical conversion.

    What test equipment is recommended to measure Bit Error Rate from Ethernet
    port on one end to Ethernet Port on the other end? Im used to using paired
    PRBS generator/receiver gear to measure BER on serial links (such as
    Fireberd 6000a), but how is this done on a packet based link? Is there
    other equipment that might provide BER or other measurements of link quality
    for this point to point link?

    Any thoughts appreciated.
    Thanks!
     
    Trs80, Jan 28, 2009
    #1
    1. Advertisements

  2. Most Ethernet gear should provide detailed stats from the interface
    itself (CRC, framing, I/O errors, underruns, overruns, etc.). If you see
    these errors increasing, you usually need to review your physical layer
    (TDR, cable scanner, etc.), or, review the traffic or resources on the
    device itself for I/O or queuing-type errors (e.g. underruns).

    TCP/IP will adjust to these errors dynamically, however, throughput will
    be severely impacted if the errors are consistent. Still, you don't
    really approach Ethernet port errors as you would a serial link
    (patterns, 1's density, and the like). Cisco and Juniper gear should
    provide enough information that you can take one of the two approaches
    mentioned.

    HTH...
     
    fugettaboutit, Jan 28, 2009
    #2
    1. Advertisements

  3. Trs80

    bod43 Guest

    Good advice already.
    You can use tools like iperf, ttcp,
    netperf, to generate traffic
    so that you can get results in a short time.
    These tools will also let you know if there are
    lost packets.

    The cisco "extended ping" is useful too.
    fping.exe for windows
    ping with "flood" option on unix.

    use large packets, 1400 bytes say.
     
    bod43, Jan 28, 2009
    #3
  4. Trs80

    Trs80 Guest

    great advise. thank you. I did find a device, the HST3000 which seems
    well suited for what I want. Im also looking for something that the less
    skilled people can use in the field and still bring back confident results.
    I dont feel that way when that have to run software apss which all its
    configuration nuances and especially interpreting data.
     
    Trs80, Jan 29, 2009
    #4
  5. Trs80

    Stephen Guest

    1 point - no Ethernet measurements at packet level can give you BER,
    since all you can tell is whether the link managed to send a packet
    error free or not.

    So - any errored packet can contain 1 or more bit errors and you
    cannot measure any closer than that.
    FWIW a lot of network errors tend to come in bursts, so measuring lost
    packets is still a useful number,

    Note such a link may include Ethernet switch or bridge elements -
    these have the annoying tendency to do their own packet checks, and
    silently "eat" corrupted packets (happens with some Ethernet NTE
    devices, SDH muxes with Ethernet support etc).
    In this case the collected error stats at the routers may not include
    any errors detected across the link.

    So - if you are really being paranoid, measure the raw fibre.

    If not you should try for something like a handheld tester, which can
    stamp every packet and give you stats about which ones didnt make it.

    Try Fluke or JDSU for some handheld testers.
    you might need 2 devices, with each sending to the other, or a remote
    acting as a mirror to swap dest + source addresses depending on how
    you want to test.

    Work has been using some JDSU handheld units that work at layer 3 and
    can inject a stream of packets on a working link at a given load
    level. A remote unit acting as a slave will "bounce" them back and the
    master measures the stats involved.
     
    Stephen, Jan 29, 2009
    #5
  6. Trs80

    dongavin123 Guest

    1 point - no Ethernet measurements at packet level can give you BER,
    since all you can tell is whether the link managed to send a packet
    error free or not.

    So - any errored packet can contain 1 or more bit errors and you
    cannot measure any closer than that.
    FWIW a lot of network errors tend to come in bursts, so measuring lost
    packets is still a useful number,

    Note such a link may include Ethernet switch or bridge elements -
    these have the annoying tendency to do their own packet checks, and
    silently "eat" corrupted packets (happens with some Ethernet NTE
    devices, SDH muxes with Ethernet support etc).
    In this case the collected error stats at the routers may not include
    any errors detected across the link.

    So - if you are really being paranoid, measure the raw fibre.

    If not you should try for something like a handheld tester, which can
    stamp every packet and give you stats about which ones didnt make it.

    Try Fluke or JDSU for some handheld testers.
    you might need 2 devices, with each sending to the other, or a remote
    acting as a mirror to swap dest + source addresses depending on how
    you want to test.

    Work has been using some JDSU handheld units that work at layer 3 and
    can inject a stream of packets on a working link at a given load
    level. A remote unit acting as a slave will "bounce" them back and the
    master measures the stats involved.
    [/QUOTE]


    Any surplus JDSU test equipment for sale?

    Don Gavin
    Test Equipment Connection Corporation
     
    dongavin123, Jun 26, 2013
    #6
    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.