Internet congestion?

Discussion in 'Broadband' started by Graham J, Dec 10, 2010.

  1. Graham J

    Graham J Guest

    I run a monitoring system that pings the external IP addresses of a few
    dozen clients, once per minute. The purpose is to warn me when any of those
    clients loses their internet connection.

    Over the last two days pings fail for many minutes at a time; yet when I
    check I can connect to the suspect clients' routers with no difficulty, and
    the routers report uptimes extending back over several days. So it would
    seem that ICMP echo packets are being dropped on a grand scale.

    This appears to relate primarily to clients who have their ADSL service from
    either BT or Sky.

    Anybody else seen this?
     
    Graham J, Dec 10, 2010
    #1
    1. Advertisements

  2. Graham J

    Graham. Guest

    I have f8lure pinging my own router, and also a VM router in Liverpool
    It is also pinging a few similar VM IPs as control specimens.
    Looking back at the last week or so graphs, if anything, things are calmer than usual.
    f8lure does have its own issues at its host end, and the sysop endlessly complains
    about them.
     
    Graham., Dec 10, 2010
    #2
    1. Advertisements

  3. Graham J

    Graham J Guest

    In the past I used F8Lure but they went off the radar a year or so back, so
    I wrote a script to run on an old PC. Nice to learn that they are back in
    action ...
     
    Graham J, Dec 10, 2010
    #3
  4. Graham J

    alexd Guest

    Meanwhile, at the uk.telecom.broadband Job Justification Hearings, Graham J
    chose the tried and tested strategy of:
    I do something similar with Smokeping, at work and at home. I haven't seen
    the pattern you're describing, but I'm not monitoring any BT ADSL [but a few
    BTnet circuits] or Sky ADSL [but a few Easynet circuits]. Is it just Sky and
    BT where you're seeing the problem?
     
    alexd, Dec 10, 2010
    #4
  5. Graham J

    Graham. Guest

    Wasn't it l8nc that dissapeared and f8lure filled the gap?
     
    Graham., Dec 10, 2010
    #5
  6. Graham J

    Graham J Guest

    You may be right - my memory is flaky. But the screens look the same ...
     
    Graham J, Dec 10, 2010
    #6
  7. Graham J

    Graham J Guest

    Indeed yes, just BT domestic services and Sky. BT Business clients seem to
    be OK.
     
    Graham J, Dec 10, 2010
    #7
  8. That's my recollection...
     
    David Quinton, Dec 11, 2010
    #8
  9. Graham J

    alexd Guest

    Meanwhile, at the uk.telecom.broadband Job Justification Hearings, Graham J
    chose the tried and tested strategy of:
    Are you graphing the response time? Is it possible that responses that take
    longer than a certain time are reported as lost? Completely dropping ICMP
    echo seems extreme to me, but then I was surprised when one ISP claimed to
    be de-prioritising ICMP during peak times, so anything's possible.
     
    alexd, Dec 11, 2010
    #9
  10. Graham J

    Chris Davies Guest

    It wouldn't surprise me if a significant percentage of ISP traffic
    was ICMP Ping from people "just checking" that servers are reachable
    from home.

    Of course, ping doesn't really show that a service is available; one
    needs a service monitor for that. And besides, if a remote box becomes
    unreachable for a few minutes, what is a (domestic) ISP realistically
    going to do about it that they wouldn't be doing based on their own
    monitoring, anyway?

    Chris
     
    Chris Davies, Dec 11, 2010
    #10
  11. That is correct
     
    Nicola Redwood, Dec 11, 2010
    #11
  12. Graham J

    Graham J Guest


    No.

    The monitoring script uses "Alive" with the default setting, so a 4-second
    timeout, and 5 retries. If all 5 tries fail, then the connection is logged
    as bad.

    Normally, any such failure is consistent with the remote router reporting
    "no connection". Very occasionally something else (possibly congestion)
    causes the failure; but this is very. infrequent.

    But last Thursday and Friday several connections (using BT and Sky) failed
    to respond to ping for long periods (tens of minutes at a time); yet I could
    connect to the routers and see that their connections had been up for many
    days.

    Maybe you're right that the ISPs are de-prioritising ICMP during peak times.

    Problem has gone away today, so it's unlikely I will be graphing the
    response times yet ...
     
    Graham J, Dec 11, 2010
    #12
  13. a TCP connection survives a temporary line break,. Only if it sends a
    keepaleve (or some other traffic) and fails to get an ack, will it
    consider the circuit to have died beyond redemption.

    Pings have much shorter timeouts. A TCP connection under 50% packet loss
    will be lumpy and slow, but will remain.

    thats about the practical limit. People start notcing speed and
    jerkiness at around 80% capacity with a few percent packet loss.

    The art of being an ISP is managing your links for over about 50%
    capacity, but zero packet loss. That's when they are profitable, and no
    one is screaming down the phone. Obviosuly every ISP with an increasing
    traffic base is monitoring support calls and booking more
    bandwidth..timed to JUST keep the dissatifaction level below complete
    desertion..
     
    The Natural Philosopher, Dec 12, 2010
    #13
    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.