[smokeping-users] Smokeping false packet loss?

[Erik Taraldsen]

Thanks for the long, good and detailed answer.  Your observations
regarding how samples and compression works explains why I'm seeng
these large losses in the aggregated graphs.  I do follow and agree on
your reasoning why this is a good thing.

This is a side track, but I'm not in complete agreement regarding that
"all loss is evil".  Losses due to transmission failures (CRC errors
etc) are bad.  Losses due to congestions are a good thing, else you
end up in a bufferbloat scenario with large latencies.  (I'm a network
engineer, one of the DSL routers had a buffer so large that we had 30
seconds delay on ADSL, which is not a good user experience).

In my scenario I do know that there is a ICMP rate limiter in the DSL
router, and at the BRAS.  Each with different points where they kick
in. And rather lots of devices on this access producing ICMP.   My
user experience is a lot better than what the graphs may indicate. :)
And I also test a lot pre production firmwares, different routers,
sometimes I swap out the ethernet cable for WLAN bridge or PLC bridge.
Sometimes we test differend profiles on the DSL link.  So, yes things
move around a lot on my setup.



-Erik