Re: Callback function instead of trace-file?

[Ahmed A-G Helmy <ahelmy@catarina.usc.edu>]

> > I'm simulating the behaviour (error recovery, congestion control) of
> > different multicast protocols (at least SRM and MFTP) in a fictive
> > large-scale network in which lots of fictive TCP-connections run. To see
> > for example how much the TCP-connections suffer from MFTP, I must observe
> > the packet loss on the various links and seperate it into TCP-connections
> > and MFTP (e.g. with the help of different flow_id's). Or I want to observe
> > potential NAK-implosions with MFTP, i.e. look at every dropped packet from
> > where it has came.
> > Further, I want to trace a repair-packet and see over which links it
> > travels in order to determine the network utilization by this particular
> > packet.
> > 
> > I'd need a custom-callback-event-handler to further evaluate a packet loss
> > or to trace a packet on it's way through the net.
> 
> One thing you should look at is the 'flow monitor'.  It is able to
> keep per-flow statistics which I think you will find useful.  One
> of these things can be associated with a particular link and will
> keep arrival/departure/drop byte+packet counts on a per-flow basis.
> If you need to introduce additional code to do things upon these
> events you could modify the in/out/drop procedures there to do what
> you like.

might I add that the callback_ var in trace.cc (class Trace) might also 
be useful there... right ?!
..
        bind("callback_", &callback_);
..
        if (callback_) {
                Tcl& tcl = Tcl::instance();
                tcl.evalf("%s handle { %s }", name(), wrk_);
        }
...

Regs,
-A
> 
> - K
>