[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ns] Packet loss on a satellite link
>
> Lloyd wrote
>
> > On Tue, 12 Jun 2001, Federico Orlandini wrote to ns-users:
>
> > I have a question: does a "$ns trace-all-satlinks $file" trace also the
> > packets dropped by a satellite link error ?
>
> Depends what you mean by a link error.
>
> I've observed packets get lost during intersatellite link handover,
> when they're on transit on a link that goes away, and those drops were
> not traced. But having the packet vanish was a Visible Sign in the
> tracefile.
>
$ns trace-all-satlinks puts tracing agents (enq, deq, drp) around the link
queues on all satellite links. If there is an error model on the receive
side (between the receive mac and link layers), then these errors
should also get traced (with an "e" character). See the procedure
Node/SatNode instproc trace-inlink-queue
in tcl/lib/ns-sat.tcl.
In general, if you want to instrument tracing on other pieces of the
simulation, you will need to add your own support for it. The
general procedure is to insert a tracing object just in front of the
drop-target of the simulation component.
Lloyd is correct that it is possible for packets to be lost in-transit.
What happens is that the channel gets attached to a different node, whose
mac address no longer matches that of the packet that was in transit, so
it is silently dropped. But making this a verbose drop results in lots of
clutter for multi-user channels.
Tom