[AG-TECH] Packet Loss

Piers O'Hanlon p.ohanlon at cs.ucl.ac.uk
Thu Aug 24 07:35:59 CDT 2006


Hi,

You could firstly just test direct unicast rat and vic sessions
between the two sites and see if you get the same packet loss. As
Marcus suggests you can use the packet info in vic and rat. You could
try altering the port numbers you use and see if it makes any
difference. Also does it happen all the time or just certain
days/time-of-day/month?

If you do see loss then you could try running traceroute from both
sides to try to see where the packets are getting lost. It will be
tricky to diagnose exactly where the problems are since you're getting
unidirectional loss but it should give you a rough idea. Under Linux
"traceroute" uses UDP packets as default - you can specify the base
port nunber with -p option. However Windows "tracert" uses ICMP
packets so you may see different results.

If you don't see loss then it maybe the unicast bridge you  were using
was being periodically overloaded.

Piers.

On 8/24/06, Chris Kendrick <kendrick at vpac.org> wrote:
> Hi All
>
> I was running an AG session between, VPAC (Melbourne, Victoria,
> Australia) and The University of Bristol in the UK Yesterday.
>
> For some reason I was getting ~5sec patches of huge packet loss (as
> shown in the RAT reception quality matrix) roughly every 3 minutes from
> them (voice went extremely choppy), but they were always receiving good
> audio from me.
>
> I was running AG2.4 compat they were running inSors. and we had met in
> the University of Bristol VV
>
> I also Started a desktop AG2.4 compat at VPAC and joined the session, I
> received packet loss from Bristol and from the original VPAC node, (due
> to the VPAC nodes both connecting to the Bristol unicast bridge, the
> traffic from the other VPAC node was obviously traveling to Bristol and
> Back and getting packet loss on the return trip)
>
> also we tried switching the Bristol inSors node between unicast and
> multicast, and we also tried going to the University of Leeds VV and
> having the Bristol node switch between unicast and multicast again.
> Still got exactly the same packet loss pattern only coming from Bristol
>
> My question is what tools tecniques can I use to further diagnose what
> is going wrong if it happens next time?
>
> Any advice gratefully received.
>
> Chris Kendrick
> VPAC
>
>




More information about the ag-tech mailing list