From owner-freebsd-stable@FreeBSD.ORG Sun Apr 9 18:11:48 2006 Return-Path: X-Original-To: freebsd-stable@freebsd.org Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 5D3E416A400 for ; Sun, 9 Apr 2006 18:11:48 +0000 (UTC) (envelope-from odix@odiisi.net) Received: from email.aon.at (warsl404pip7.highway.telekom.at [195.3.96.91]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9DDA243D45 for ; Sun, 9 Apr 2006 18:11:47 +0000 (GMT) (envelope-from odix@odiisi.net) Received: (qmail 20792 invoked from network); 9 Apr 2006 18:11:45 -0000 Received: from m1311p012.adsl.highway.telekom.at (HELO odi45.odihome.lan) ([80.121.35.204]) (envelope-sender ) by smarthub73.highway.telekom.at (qmail-ldap-1.03) with SMTP for ; 9 Apr 2006 18:11:45 -0000 From: odi To: freebsd-stable@freebsd.org Content-Type: text/plain Date: Sun, 09 Apr 2006 20:11:44 +0200 Message-Id: <1144606304.26234.32.camel@localhost> Mime-Version: 1.0 X-Mailer: Evolution 2.4.1 Content-Transfer-Encoding: 7bit Subject: network pertfomance low with RELENG_6_0 from 2nd of march X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 09 Apr 2006 18:11:48 -0000 Hi there, note: I've send this email a couple of days ago, but it seems to be hanging around in a mail server ... I'm relatively new to freebsd (since 6.0rc), but not new to unix (first contact with unix was xenix286), since year 2000 working with gnu/linux (rock, gentoo). the machine is a pentium3 with 600MHz(or 650MHz?) and 256MB of ram, make.conf: newfw# cat /etc/make.conf CPUTYPE?=pentium3 CFLAGS= -O2 -pipe # added by use.perl 2006-03-10 14:11:48 PERL_VER=5.8.8 PERL_VERSION=5.8.8 now the problem: simple file transfer using scp screw up cpu load to near 100% by only using around 50% (around 5MB/s) of the bandwidth (100Mbit switched network), using http protocol with apache2 bandwidth utilisation increase to nearly 75% (around 7MB/s), ok scp is more cpu intensive. I've tried different settings, switch to device polling, reduce timer to 250Hz, switch on and of inflight, with only minimal difference in throughput. I've also tried different nics (xl, rl, sis) with little or now difference in performance, so I'm a little bit lost :-( With linux I've reached the pyhsical limits of the network with bot scp and http with much less cpu utilisation, so where are the correct "knobs" to adjust the system :-). I don't want to start a flame war and I like the network features of freebsd, so I want to use it for this kind of machines, but without the performance lack. regards manfred (odi)