From owner-freebsd-smp@FreeBSD.ORG Sat Sep 2 09:47:24 2006 Return-Path: X-Original-To: freebsd-smp@freebsd.org Delivered-To: freebsd-smp@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 0B56516A4EC for ; Sat, 2 Sep 2006 09:47:24 +0000 (UTC) (envelope-from rwatson@FreeBSD.org) Received: from cyrus.watson.org (cyrus.watson.org [209.31.154.42]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9D05843E2F for ; Sat, 2 Sep 2006 09:46:53 +0000 (GMT) (envelope-from rwatson@FreeBSD.org) Received: from fledge.watson.org (fledge.watson.org [209.31.154.41]) by cyrus.watson.org (Postfix) with ESMTP id 059B246B3E; Sat, 2 Sep 2006 05:46:31 -0400 (EDT) Date: Sat, 2 Sep 2006 10:46:30 +0100 (BST) From: Robert Watson X-X-Sender: robert@fledge.watson.org To: kumaresh pandian In-Reply-To: <20060823050520.47768.qmail@web8604.mail.in.yahoo.com> Message-ID: <20060902104213.C84468@fledge.watson.org> References: <20060823050520.47768.qmail@web8604.mail.in.yahoo.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: freebsd-smp@freebsd.org Subject: Re: Netpref results on SMP machine X-BeenThere: freebsd-smp@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: FreeBSD SMP implementation group List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 02 Sep 2006 09:47:24 -0000 On Wed, 23 Aug 2006, kumaresh pandian wrote: > I ran the netperf test of TCP_STREAM for 35 hours on a FreeBSD 6.1 SMP > machin and a FreeBSD Single Processor machine which has the following > configurations There have been a number of reports of significantly reduced loopback IP performance on 6.x as a result of preemption, which can result in context switch thrashing on some systems. There has been some amount of discussion of this problem, in particular at the recent BSDCan developer summit, and there are tentative plans for a fix. You might want to try disabling preemption in your kernel configuration to see what the impact is. However, as you're not doing side-by-side comparisons in your tests, it's hard to compare the results in order to decide what problem may exist. Robert N M Watson Computer Laboratory University of Cambridge