From owner-freebsd-net Fri Aug 4 15:17: 0 2000 Delivered-To: freebsd-net@freebsd.org Received: from bmah-freebsd-0.cisco.com (bmah-freebsd-0.cisco.com [171.70.84.42]) by hub.freebsd.org (Postfix) with ESMTP id 3FAC737BA04 for ; Fri, 4 Aug 2000 15:16:55 -0700 (PDT) (envelope-from bmah@cisco.com) Received: (from bmah@localhost) by bmah-freebsd-0.cisco.com (8.11.0/8.11.0) id e74MGMk14703; Fri, 4 Aug 2000 15:16:22 -0700 (PDT) (envelope-from bmah) Message-Id: <200008042216.e74MGMk14703@bmah-freebsd-0.cisco.com> X-Mailer: exmh version 2.2 06/23/2000 with nmh-1.0.4 To: Garrett Wollman Cc: bmah@cisco.com, freebsd-net@FreeBSD.ORG Subject: Re: cvs commit: src/sys/netinet tcp_output.c In-Reply-To: <200008042131.RAA33779@khavrinen.lcs.mit.edu> References: <200008042014.QAA05452@cholla.INRS-Telecom.UQuebec.CA> <200008042107.e74L7Tu13995@bmah-freebsd-0.cisco.com> <200008042131.RAA33779@khavrinen.lcs.mit.edu> Comments: In-reply-to Garrett Wollman message dated "Fri, 04 Aug 2000 17:31:44 -0400." From: bmah@cisco.com (Bruce A. Mah) Reply-To: bmah@cisco.com X-Face: g~c`.{#4q0"(V*b#g[i~rXgm*w;:nMfz%_RZLma)UgGN&=j`5vXoU^@n5v4:OO)c["!w)nD/!!~e4Sj7LiT'6*wZ83454H""lb{CC%T37O!!'S$S&D}sem7I[A 2V%N&+ X-Image-Url: http://www.employees.org/~bmah/Images/bmah-cisco-small.gif X-Url: http://www.employees.org/~bmah/ Mime-Version: 1.0 Content-Type: multipart/signed; boundary="==_Exmh_-2095522332P"; micalg=pgp-sha1; protocol="application/pgp-signature" Content-Transfer-Encoding: 7bit Date: Fri, 04 Aug 2000 15:16:22 -0700 Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org --==_Exmh_-2095522332P Content-Type: text/plain; charset=us-ascii If memory serves me right, Garrett Wollman wrote: > < > > A RED queue would drop a random packet from the full queue. > > Actually, no. A RED queue would drop a packet as it is enqueued, at > random intervals which depend on the current length of the queue. > Many people get this wrong when the idea is first explained to them; I > know I did. Read the paper for a full explanation of how this works.[1] Mea culpa, thanks for the correction. It's been too long since I read the paper. If it makes you feel any better, I have *nothing* to do whatsoever with the RED implementation in any Cisco product. :-p > > If you can get those flows to back off (e.g. TCP congestion > > control), > > Specifically, I think the problem with Archie's patch is that it might > result in TCP not backing off at all. I think I see your concern but I am not sure whether or not this is true. Given that I started off my previous email with a totally false statement, I'm not sure if I should venture out on a limb again, but: From the commit log, the error case Archie is concerned with is if ip_output returns with ENOBUFS. If this is true, then after Archie's patch gets executed, the code below the out: label will call tcp_quench(), which closes the congestion window down to one segment. The fact that the patch already reset tp->snd_nxt doesn't change that. > RED doesn't deal well with > unresponsive flows. Agreed. Bruce. --==_Exmh_-2095522332P Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: PGPfreeware 5.0i for non-commercial use MessageID: HuZKOCkYKe88R65elv+ql4bPbwEyn4xc iQA/AwUBOYtAttjKMXFboFLDEQJlkwCfTiMVV4cCrBWxlUDLG+vtRwHq1yoAoJAb wbp/u8hqqDGPdcSbe8NpXDhQ =d0l+ -----END PGP SIGNATURE----- --==_Exmh_-2095522332P-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message