From owner-freebsd-net Thu Aug 26 21:30:28 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 5FCD514F64 for ; Thu, 26 Aug 1999 21:30:25 -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 AAA04468; Fri, 27 Aug 1999 00:30:21 -0400 (EDT) (envelope-from wollman) Date: Fri, 27 Aug 1999 00:30:21 -0400 (EDT) From: Garrett Wollman To: net@FreeBSD.org Subject: Call for Review: TCP timer changes In-Reply-To: <199908270411.AAA04392@khavrinen.lcs.mit.edu> Message-Id: <199908270419.AAA04425@khavrinen.lcs.mit.edu> References: <19990826224556.04405@right.PCS> <199908270411.AAA04392@khavrinen.lcs.mit.edu> Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org [Resending to -net after typo fix...] < I'm particularly interesting in hearing from Alpha users. Another area which might need further investigation is better tuning of ncallout and callwheelsize. Right now, the callout wheel is sized on the assumption that there will only ever be `ncallout' callouts active simultaneously. (This is not a correctness constraint, but affects efficiency.) This means that the most efficient operation for machines with large numbers of outstanding TCP connections will likely require a partial decoupling of the two; since TCP allocates its own callout structures, high values of ncallout would needlessly waste memory. -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