Date: Tue, 24 Apr 2001 21:03:32 -0700 From: Brooks Davis <brooks@one-eyed-alien.net> To: Mark <mark@offworld1.net> Cc: freebsd-small@FreeBSD.ORG Subject: Re: Is BSD capable?? Message-ID: <20010424210332.A30795@Odin.AC.HMC.Edu> In-Reply-To: <01042503542301.15147@offworld1.net>; from mark@offworld1.net on Wed, Apr 25, 2001 at 03:50:33AM -0400 References: <002001c0cc6d$be9aee20$8995efd1@quick> <01042421341401.14955@offworld1.net> <20010424145138.B30762@Odin.AC.HMC.Edu> <01042503542301.15147@offworld1.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--AqsLC8rIMeq19msA
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Wed, Apr 25, 2001 at 03:50:33AM -0400, Mark wrote:
> What I have here is a network set up something like the following (simpli=
fied
> the best I can). Here at the NOC I have a half duplex Access Point (AP1).
> This is linked to Remote Site 1 (RS1). Connected to RS1 are 2 more Access
> Points (AP2, AP3). All are half duplex. End users connect to AP2 and AP3.=
What
> happens is, that when a user connects to say AP2 and starts a download, a=
nd a
> user connected to AP3 attempts to send an MP3 to his bud up in Spider Bre=
ath
> MO., AP2 and AP3 both attempt to access the gateway, RS1, in two separate
> directions at the same time into a half duplex back haul link.=20
>=20
> In practice, one or the other suffers, and when you have 3 doing a downlo=
ad=20
> and one fighting upstream, guess which one loses. That is because there is
> no control between AP2 and AP3. Neither one knows what the other is doing.
> Being able to install a full duplex link in the backbone would give me a
> big boost in overall network performance. Trouble is that a full duplex
> link is thou$and$ of dollars more than 2 half duplex links. If this can
> be done, great, If not, I start saving my pennies.=20
Ok, so if I understand correctly, you have an oversubscribed half duplex
link that gets pounded by collisions when in heavy (bi-directional) use.
If that's the case, you might try cranking the rts threshold way down
on both ends of the link (that's the -r option to {wi,an}control).
This introduces a bit more latency by forcing both ends to ask nicely
before they send anything, but if most of you're traffic is big it could
be a hugh win. Enabling it on my access point raised ping latench from
1.9ms to 2.1ms, enabling it on my Cisco 340 card turned off the pipe so it
appears there's some work to be done in that driver. ;-) If that works,
then you could use the money you would have spent building a full-duplex
link to double your link speed. ;-)
-- Brooks
--=20
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4
--AqsLC8rIMeq19msA
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.4 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iD8DBQE65kyKXY6L6fI4GtQRAjthAJwKC6qvrDpyl6IO6EWmMFEvzU4K/gCgubM+
lLvrvFbKKdGXvqpSw/yAKrU=
=m4A/
-----END PGP SIGNATURE-----
--AqsLC8rIMeq19msA--
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-small" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010424210332.A30795>