RE: [ns] goodput calculation

[Michael Welzl <michael@tk.uni-linz.ac.at>]

After recommendations from this list, I finally switched from hacking
ns and using bound variables to parsing the trace file ... and it works
MUCH better for me.

Of course, trace files from "$ns trace-all" can become quite large, but
if you pipe them directly to a perl script instead of saving and parsing
afterwards, it works just fine.

Cheers,
Michael

> -----Original Message-----
> From: owner-ns-users@ISI.EDU 
> [mailto:owner-ns-users@ISI.EDU]On Behalf Of
> Lloyd Wood
> Sent: Wednesday, June 13, 2001 2:00 PM
> To: Rafal Wielicki
> Cc: ns-users mailing list
> Subject: Re: [ns] goodput calculation
> 
> 
> On Wed, 13 Jun 2001, Rafal Wielicki wrote:
> 
> > Hi Lloyd !
> > 	How can I modify an example form your homepage to 
> extract goodput for 
> > each TCP source-sink pair ?
> 
> You'll need to hack ns (C++ and Tcl) to create a bound variable at the
> sink, which gets incremented in the sink's recvBytes() function (or
> before that is called). Once that's done, you can write OTcl trace
> procedures for each sink, recording the value of the sink's variable
> every so often, just as you do for cwnd_ at each sender.
> 
> At least you'll only have to write this once for all types of
> TCPSink... I'd worry about resets at the sender too.
> 
> L.
> 
> and your TCP simulations will slow to a crawl.
> 
> <L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>
>