Hi,


Am Mi, 18.07.12 um 17:49:35 Uhr
schrieb Antony Antony <antony@ripe.net>:

>
> On Wed, Jul 18, 2012 at 11:48:08AM +0000, Bartsch, Rene wrote:
> > Hi,
> >
> > I've just started hosting a probe and I'm new to the list, so a "Hello" to all! :)
> >
> > Does the probe send simple pings or does it send flood pings (e.g. ping -f -c 1000 <destination>) for detection of packet loss?
> >

If you would do a flood ping it is more likely you will observe congestion
in the LAN when the probe is connected to a slow uplink. More interesting
for WAN measurements is the loss under 'normal' conditions. Consider the
probe beeing installed at a residential DSL or cable line, where one of
the probes I host ist located.

If I start a huge download I will have congestion on my upstream because
of the ACKs I send. A probe would now observe huge loss when doing a flood ping.

I've installed the first probe (just registered for a second) on a residential socket on today noon. It shows packet loss for c.root-servers.net (192.33.4.12) from time to time and continous packet loss of 100% for m.root-servers.net (202.12.27.3). Netalyzer-Tests show a packet loss of 8-17% for the last two weeks (problems habe been around for about half a year, getting more and more annyoing for the last three month), while connections inside Germany seem to work for the users meanwhile. The 16/1-MBit/s last mile has one or zero non-correctable errors in fifteen minutes (DSL). CPE (IAD 3222 -> IAD 3221 -> Fritzbox 7390) and wiring has been changed completely. For me, it looks like a broken or misconfigured router in the ISPs backbone, but they keep insisting everything is fine with their backbone. :(

When I did simple ping tests, no packet loss was shown on the unused line, but when using a flood ping with 1000 packets I can see packet loss. As I'm no forensic scientist I'm trying to find a way to localize the problem ...