Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 7 Nov 2004 14:04:59 +0100
From:      Pawel Malachowski <pawmal-posting@freebsd.lublin.pl>
To:        ipfw-mailings <freebsd-ipfw@freebsd.org>
Subject:   Re: Dummynet dynamically assigned bandwidth
Message-ID:  <20041107130459.GB75933@shellma.zin.lublin.pl>
In-Reply-To: <1099831431.652.43.camel@Mobile1.276NET>
References:  <1099819314.652.13.camel@Mobile1.276NET> <20041107094433.GA56141@shellma.zin.lublin.pl> <1099822179.652.18.camel@Mobile1.276NET> <20041107103742.GA74864@shellma.zin.lublin.pl> <1099831431.652.43.camel@Mobile1.276NET>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Nov 07, 2004 at 03:43:51PM +0300, Martes Wigglesworth wrote:

> I am so sorry to have to demonstrate my ignorance on this subject,
> however, I am not quite sure what you mean by 50*1.5KB/s/16KB/s. Where
> is that rational expression comming from?  Also, since I really have no
> good context for understanding your possible solution, could you supply
> more info, or an example?

Your example:
pipe 1 config bw 128Kbit/s
queue 1 config pipe 1 config mask dst-ip 0xffffffff
queue 1 ip from any to ${int_net} in recv ${ext_dev}

128Kbit/s = 16KBytes/s
As described in ipfw(8), default size of queues and pipes is 50 slots.
This may valid for 10Mbit/s, but not for 128Kbit/s...

Proposal:
pipe 1 config bw 128Kbit/s queue 4KB
queue 1 config pipe 1 mask dst-ip 0xffffffff queue 3KB
queue 1 ip from any to ${int_net} in recv ${ext_dev}

Lowering size of queues and pipes from 50 slots (which can give 75KB
with 1500B MTU) to 2 slots (or 3-4KBytes, I prefer specifying sizes
in bytes) will help TCP in `smooth' working by lowering maximum delay.
Just give it a try.

See also:
http://groups.google.pl/groups?selm=blc6oa%2416bm%241%40FreeBSD.csie.NCTU.edu.tw


-- 
Paweł Małachowski



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041107130459.GB75933>