Date: Sun, 5 Nov 2000 17:57:22 -0800 From: Kris Kennaway <kris@FreeBSD.ORG> To: "Richard A. Steenbergen" <ras@e-gerbil.net> Cc: David Greenman <dg@root.com>, freebsd-net@FreeBSD.ORG Subject: Re: tcp sendspace/recvspace Message-ID: <20001105175722.A8886@citusc17.usc.edu> In-Reply-To: <Pine.BSF.4.21.0011052035270.306-100000@overlord.e-gerbil.net>; from ras@e-gerbil.net on Sun, Nov 05, 2000 at 08:36:14PM -0500 References: <20001105172400.A8778@citusc17.usc.edu> <Pine.BSF.4.21.0011052035270.306-100000@overlord.e-gerbil.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--zhXaljGHf11kAtnf Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sun, Nov 05, 2000 at 08:36:14PM -0500, Richard A. Steenbergen wrote: > On Sun, 5 Nov 2000, Kris Kennaway wrote: >=20 > > On Sun, Nov 05, 2000 at 02:02:44PM -0800, David Greenman wrote: > > > I've been messing around with the net.inet.tcp.sendspace and=20 > > > net.inet.tcp.recvspace parameters on ftp.freesoftware.com and have fo= und > > > that there is a significant performance improvement when increasing t= hese > > > to 32768 bytes. Apparantly there are enough systems out there with hi= gher > > > window maxes that it really does make a difference. By significant=20 > > > improvement, I mean about a average of a 20% increase in Mbps per use= r, > > > and this was just the change over a 30 minute period with lots of con= nections > > > still using the old 16K values. > > > Any objections to increasing the defaults in FreeBSD to 32K? > >=20 > > Won't this impact the latency of multiple connections over a slow PPP l= ink? >=20 > Thats what the congestion window is for. Any limitation of transmits based > on the advertised memory limitation window is totally improper. Perhaps it was a bug, but I used to see e.g. FTP transfers which were running at full speed totally monopolizing my modem bandwidth (then a 14.4k), and other sessions not being able to receive their "fair share". Tweaking net.inet.tcp.recvspace to give only a second or two worth of data transfer reduced the latency to acceptable levels. Maybe this has been fixed by now - I haven't noticed it since I upgraded to a 56k modem. I'll try increasing my system to 32768 and see if it has any effect. Kris --zhXaljGHf11kAtnf Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.4 (FreeBSD) Comment: For info see http://www.gnupg.org iEYEARECAAYFAjoGEAIACgkQWry0BWjoQKXcwQCgq/gqLgX+Ww9vFeIvniPG6Zwi ROoAoL+42nSzUBaCTnHWVpOTJb5lugyQ =CG2d -----END PGP SIGNATURE----- --zhXaljGHf11kAtnf-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20001105175722.A8886>