From owner-cvs-all Thu Jul 18 12:29:30 2002 Delivered-To: cvs-all@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id DBDAE37B400; Thu, 18 Jul 2002 12:29:22 -0700 (PDT) Received: from apollo.backplane.com (apollo.backplane.com [216.240.41.2]) by mx1.FreeBSD.org (Postfix) with ESMTP id 8B90543E7B; Thu, 18 Jul 2002 12:29:22 -0700 (PDT) (envelope-from dillon@apollo.backplane.com) Received: from apollo.backplane.com (localhost [127.0.0.1]) by apollo.backplane.com (8.12.5/8.12.4) with ESMTP id g6IJTMCV019579; Thu, 18 Jul 2002 12:29:22 -0700 (PDT) (envelope-from dillon@apollo.backplane.com) Received: (from dillon@localhost) by apollo.backplane.com (8.12.5/8.12.4/Submit) id g6IJTMsF019578; Thu, 18 Jul 2002 12:29:22 -0700 (PDT) (envelope-from dillon) Date: Thu, 18 Jul 2002 12:29:22 -0700 (PDT) From: Matthew Dillon Message-Id: <200207181929.g6IJTMsF019578@apollo.backplane.com> To: Bosko Milekic Cc: cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: src/sys/netinet tcp_subr.c tcp_timer.c tcp_timer.h References: <200207181906.g6IJ6CJI048667@freefall.freebsd.org> <20020718150731.A61071@unixdaemons.com> Sender: owner-cvs-all@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG :On Thu, Jul 18, 2002 at 12:06:12PM -0700, Matt Dillon wrote: :[...] :> Please note that the 200ms slop is debateable. :> This commit is not meant to be a line in the sand, :> and if the community winds up deciding that increasing :> it is the correct solution then it's easy to do. :> Note that larger values will destroy performance :> on lossy networks while smaller values may result in :> a greater number of unnecessary retransmits. : : Thanks! : :> Revision Changes Path :> 1.137 +3 -1 src/sys/netinet/tcp_subr.c :> 1.52 +8 -0 src/sys/netinet/tcp_timer.c :> 1.21 +21 -4 src/sys/netinet/tcp_timer.h : :Cheers, :-- :Bosko Milekic :bmilekic@unixdaemons.com This will at least make it easier for people to test different combinations. Note that net.inet.tcp.rexmit_slop is instantaniously applied, but net.inet.tcp.rexmit_min is cached when a new TCP connection is initiated and so you must exit and re-create the TCP connection when testing changes in rexmit_min (e.g. you have to quit out of ftp and ftp back in, if using passive-mode ftp). Recommended testing (transmission path from the modified host, not the receive path): * over a wireless network * over dummynet with packet losses of 0.1% to 10% (0.001 to 0.10). * over a LAN * over the internet (cablemodem, DSL, etc...) * with window scaling and large buffers Test with slop values of 20, 50, 100, 200, 500, and 1000 ms. Observe netstat -s | head -20 output for number of retransmits sent and duplicate acks received. Observe performance variances with netstat -in 1, observe interactive performance typing over ssh through dummynet (or on a wireless network at the edge of its range). etc... -Matt Matthew Dillon To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message