Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 17 Mar 2004 03:32:06 -0800
From:      Luigi Rizzo <luigi@iet.unipi.it>
To:        Karim Fodil-Lemelin <kfl@xiphos.ca>
Cc:        freebsd-net@freebsd.org
Subject:   Re: Dummynet Limitations
Message-ID:  <20040317033206.A23760@xorpc.icir.org>
In-Reply-To: <40575A8F.8090402@xiphos.ca>; from kfl@xiphos.ca on Tue, Mar 16, 2004 at 02:50:39PM -0500
References:  <40575A8F.8090402@xiphos.ca>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Mar 16, 2004 at 02:50:39PM -0500, Karim Fodil-Lemelin wrote:
> Hi
> 
>     This code:
> 
> <code>
>      if (pipe.delay > 10000)
>                             errx(EX_DATAERR, "delay %d must be < 10000", 
> pipe.delay);
> </code>
> 
> in /usr/src/sbin/ipfw/ipfw.c
> 
> Limits the pipe delay for dummynet to 10 seconds. Is there any reason 
> for this? Also, no such limit is imposed on the bandwidth why?
> Memory (amount of mbufs/mbclusters) is obviously a limit here but I was 
> wondering if something else was hidden in this statement.

back in 1996 when i wrote that line of code, my test machine had
16MB of RAM and i did not want to exhaust it with a poorly
configured pipe. You are certainly welcome to increase it
or even better make it into a sysctl variable.

But keep in mind that pipe.delay is the propagation delay, and i
cannot make any sense (except for simulating deep-space communications)
of a larger delay, in which case, though, communication is
so peculiar that almost surely you need a different test
environment, not something IP-based as dummynet.
Also, in such cases, you probably want to simulate things
faster than realtime -- makes no sense to send a packet
and wait hours for it to be delivered.

cheers
luigi



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040317033206.A23760>