From owner-freebsd-isp Wed Mar 20 12:16:24 1996 Return-Path: owner-isp Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id MAA00778 for isp-outgoing; Wed, 20 Mar 1996 12:16:24 -0800 (PST) Received: from Root.COM (implode.Root.COM [198.145.90.17]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id MAA00751 for ; Wed, 20 Mar 1996 12:16:16 -0800 (PST) Received: from localhost (localhost [127.0.0.1]) by Root.COM (8.6.12/8.6.5) with SMTP id MAA03338; Wed, 20 Mar 1996 12:14:40 -0800 Message-Id: <199603202014.MAA03338@Root.COM> X-Authentication-Warning: implode.Root.COM: Host localhost didn't use HELO protocol To: Joe Greco cc: nate@sri.MT.net (Nate Williams), freebsd-isp@freefall.freebsd.org Subject: Re: Microsoft "Get ISDN"? In-reply-to: Your message of "Wed, 20 Mar 1996 12:19:40 CST." <199603201819.MAA29896@brasil.moneng.mei.com> From: David Greenman Reply-To: davidg@Root.COM Date: Wed, 20 Mar 1996 12:14:40 -0800 Sender: owner-isp@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >> > I am curious, why do you use SLIP for your dedicated connections? >> >> I don't speak for Joe, but I use SLIP (w/VJ compression) whenever >> possible since it uses less overhead and I seem to have lower latency >> and higher throughput than using both user-mode and kernel-mode PPP on >> FreeBSD boxes. > >I don't have anything to say about higher throughput since in my experience >it's only a mild difference, but the latency issue is mainly due to ppp's >default 1500 mtu. Lowering that (I know one fella who uses 296) will help >latency issues quite a bit. With 1500, you only get two or three packets >per second through the link if somebody is running a large transfer of some >sort. > >Most sites which run dedicated connections will have multiple people using >the link simultaneously, so the lower mtu gives the impression of faster >response. This hurts overall throughput mildly, buuuuut there's always a >tradeoff to be made. > >fyi: SLIP uses a 552 mtu. While this is true when using traditional modems, it is far less true when using modern modems with data compression. The modems themselves buffer a potentially large amount of data (1Kb or more of compressed data), and this will often make small MTUs perform worse in all respects because of the higher packet overhead. I know this through direct experimentation with different MTU values on several different 28.8K modems and the results of that are how I came up with the 552 value that we are currently using in FreeBSD. -DG David Greenman Core-team/Principal Architect, The FreeBSD Project