Another alternative is up_rtt.monitor. It uses UDP echo, but fails over to
TCP echo if none of the UDP packets are returned. Logging this information
can tell you a lot about the health of a circuit over time.

Some early examples are shown in:

http://www.usenix.org/publications/library/proceedings/lisa98/full_papers/meek/meek.pdf


Counting the dropped packets provides an estimate of packet loss on the
circuit as well.

The latest version is available at:

 http://wanpcap.sourceforge.net/mon/

If you are going to log, you should use this newer version because it logs
the raw RTTs for better statistical treatment.

Jon

On Fri, 18 Oct 2002, Stephane Bortzmeyer wrote:

> On Fri, Oct 18, 2002 at 12:05:26PM +0200,
>  Baco <[EMAIL PROTECTED]> wrote
>  a message of 132 lines which said:
>
> > When we are monitoring a leased line and the traffic is running very high the 
>router or the modem
> > drop the ICMP requests so the leased line apears to be down but it is still 
>running. One solution is
> > maybe to use TCP Ping as an alternative to ICMP echo.
>
> I agree :-) http://echoping.sourceforge.net/
>
> > Is it possible to include the following patch or a derivate in the next release of 
>Mon ?
>
> No! Adding a new monitor is fine, changing an existing monitor to
> something else, with a different semantic is not.
>
> No probe is perfect. fping has its problems, echoping too and tping
> too. We need to have the choice.

_______________________________________________
mon mailing list
[EMAIL PROTECTED]
http://linux.kernel.org/mailman/listinfo/mon

Reply via email to