Re: [ns] trace file for target flow

[Tarik Alj <aljtarik@cholla.inrs-telecom.uquebec.ca>]

You would want to have a fast CPU and some memory because this is going to slow 
down your simulation (e.g. it *might* take an awful lot of time to run), if I 
may add to Lloyd's suggestion.

>Date: Thu, 13 Jul 2000 16:27:07 +0100 (BST)
>From: Lloyd Wood <l.wood@eim.surrey.ac.uk>
>X-Sender: eep1lw@petra.ee.surrey.ac.uk
>To: Chengzhi Li <chengzhi@rice.edu>
>cc: ns-users@mash.cs.berkeley.edu
>Subject: Re: [ns] trace file for target flow
>X-url: http://www.ee.surrey.ac.uk/Personal/L.Wood/
>X-no-archive: yes
>MIME-Version: 1.0
>
>On Wed, 12 Jul 2000, Chengzhi Li wrote:
>
>> I run a simulation for 45Mpbs links and more than 4000 VoIP
>> flows in a 6 nodes topology over 60 min.  I only need to know 
>> the performance about target flows (only 80 flows).  Although 
>> I use trace-queue {n1 n2 file} to minimize the trace file, it 
>> is still huge.  Is it possible to only store the entries related
>> to target flows?
>
>pipe your output through a pattern matcher that only matches lines
>relating to target flows. This trick for compressing tracefiles is
>well-known:
>
>set f [open "|gzip -c > outputtrace.tr.gz" w]
>
>you could instead write e.g. an awk script that only writes out lines
>if e.g. $8, $9 and $10 match your requirements for a flow, or do that
>with a perl WHILE (<OUTFILE>) loop and regexp of some sort...
>
>set f [open "|awk -f strip.awk |gzip -c > outputtrace.tr.gz" w]
>
>or
>
>set f [open "|perl strip.pl |gzip -c > outputtrace.tr.gz" w]
>
>L.
>
><L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>

Tarik