From owner-freebsd-net Fri May 28 19:51:15 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 D956114EED for ; Fri, 28 May 1999 19:51:10 -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 WAA24826; Fri, 28 May 1999 22:50:54 -0400 (EDT) (envelope-from wollman) Date: Fri, 28 May 1999 22:50:54 -0400 (EDT) From: Garrett Wollman Message-Id: <199905290250.WAA24826@khavrinen.lcs.mit.edu> To: David Gilbert Cc: freebsd-net@FreeBSD.ORG Subject: Linux and Solaris tcp beating FreeBSD 4:1 In-Reply-To: <14159.19240.129931.14715@trooper.velocet.ca> References: <14159.19240.129931.14715@trooper.velocet.ca> Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org < said: > I havn't done a packet dump yet... but why is Linux beating FreeBSD > 4:1 and what can I do about it? You don't say anything about the latency of the link, but here's my guess: The TCP retransmit timer is initialized to min(2, srtt + 4*rttvar). If your link is lossy, you'll probably be in retransmit a lot. Unfortunately, the TCP timers in -current are measured in 1/2-second intervals (a remnant from the VAX implementation), which means that the minimum retransmit timeout in FreeBSD is one second. This is probably an order of magnitude greater than the round-trip time. Linux and Solaris use different TCP timer implementations which have much higher-resolution timers. I have some work in progress which will fix this, and allow us to have retransmit timeouts measured in timer ticks. Unfortunately, it needs about 12 hours more work and I don't have the time or resources for that right now. (I'm also waiting on Warner to finish the PC-Card de-klugery so that I can update my laptop to -current again.) -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