From owner-freebsd-net Tue Jul 6 6:46:23 1999 Delivered-To: freebsd-net@freebsd.org Received: from khavrinen.lcs.mit.edu (khavrinen.lcs.mit.edu [18.24.4.193]) by hub.freebsd.org (Postfix) with ESMTP id D419214F2B for ; Tue, 6 Jul 1999 06:46:21 -0700 (PDT) (envelope-from wollman@khavrinen.lcs.mit.edu) Received: (from wollman@localhost) by khavrinen.lcs.mit.edu (8.9.1/8.9.1) id JAA29874; Tue, 6 Jul 1999 09:46:19 -0400 (EDT) (envelope-from wollman) Date: Tue, 6 Jul 1999 09:46:19 -0400 (EDT) From: Garrett Wollman Message-Id: <199907061346.JAA29874@khavrinen.lcs.mit.edu> To: dg@root.com Cc: net@FreeBSD.ORG Subject: Re: mbuf-less IP? In-Reply-To: <199907052053.NAA05826@implode.root.com> References: <3780DE74.2DD4F526@greycat.com> <199907052053.NAA05826@implode.root.com> Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org < said: > We've made substantial improvements in FreeBSD to the way that mbufs are > allocated and used which have mitigated much of the old problems. It could > still be a lot better, however. It's really the socket layer that needs the > rewrite; I don't think the TCP/IP stack itself would be that difficult. Actually, the principal notion of Van's work is to completely eliminate the socket layer entirely, and write each protocol to directly interface to the user (well, to the back end of the system calls). One of the significant barriers to performance he identified was the presence of two separate flow-control models which operate essentially as `ships in the night' and often interfere with each other. (That's why the write size/transmit speed curve is discontinuous.) -GAWollman -- Garrett A. Wollman | O Siem / We are all family / O Siem / We're all the same wollman@lcs.mit.edu | O Siem / The fires of freedom Opinions not those of| Dance in the burning flame MIT, LCS, CRS, or NSA| - Susan Aglukark and Chad Irschick To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message