Hello Rob.
On Thursday, 04 March, 2021, you wrote
> One of my machines was showing 8% packet loss
> when pinging the same site as another machine on
> the same hub at the same time which was
> reporting 0%.
...
> What I find odd though is why not 0% or 100%?
> Surely the wires inside the cable can only break
"All signals are analogue, even if they're
digital."
The ethernet cable is a transmission line,
carrying high-frequency signals from the
transmitter to the receiver. Although 100BaseTx/
1GBaseT use twisted-pair, it's little different
from the coaxial cable connecting a television or
radio to its antenna.
Without attaching the cable to a time-domain
reflectometry scope, it's not possible to say what
caused the packet loss. However:
- A fracture in the cable or in an RJ-45/conductor
joint will block DC, but not the high-frequency
AC of the ethernet signal. To the latter, the
fracture will look like a low-value capacitor;
the signal will be degraded, but not blocked
completely.
- Ethernet uses forward error correction (FEC).
Extra bits are added to the transmitted data
stream, adding redundancy to the data. At the
receiver, these bits are used to detect
corrupted data and, if there are not too many,
recover lost bits. This means that ethernet can
tolerate a certain amount of interference before
you start seeing packet loss at the ICMP layer.
Damage to the cable can make it more susceptible
to interference - you will not notice until the
bit errors exceed the ability of the FEC to
correct it.
For information on how this sort of problem is
measured, have a look at:
https://en.wikipedia.org/wiki/Eye_diagram
> it wasn't as if the cable was being moved
> around - it was stationary. To lose one packet
> every 12 or so seems very odd for a cable issue.
Even if the cable wasn't being moved deliberately,
it was still being influenced by temperature
changes. The resulting expansion and contraction
of the copper and PVC can cause a dodgy joint to
break, or a weak point to fracture.
> would the issue have been to do with:
>
> a) the length of the cable or
> b) the quality of the cable or
> c) both?
Most likely (b). Well-manufactured cables with
good strain relief will still fail eventually, but
they will last longer.
For a short while, I worked for an ISP. A regular
task was to check the ethernet interface
statistics on the routers and switches for rapidly
increasing FEC or CRC errors. That information
would allow us to detect dodgy cables and
interfaces that were on their last legs.
Nick.
--
Nick Chalk ................. once a Radio Designer
Confidence is failing to understand the problem.
--
Please post to: Hampshire@???
Web Interface:
https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL:
http://www.hantslug.org.uk
--------------------------------------------------------------
