From owner-freebsd-current@FreeBSD.ORG Sat Jul 17 19:59:37 2004 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 2503A16A4CE for ; Sat, 17 Jul 2004 19:59:37 +0000 (GMT) Received: from mail.omut.org (mail.omut.org [216.218.215.140]) by mx1.FreeBSD.org (Postfix) with ESMTP id F1FFD43D1F for ; Sat, 17 Jul 2004 19:59:36 +0000 (GMT) (envelope-from lxv@omut.org) Received: from tadpole.intranet (mix-anchor.intranet [10.10.10.251]) by mail.omut.org (8.12.11/8.12.11) with ESMTP id i6HJxXcI069935 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Sat, 17 Jul 2004 12:59:34 -0700 (PDT) (envelope-from lxv@omut.org) Received: from tadpole.intranet (localhost [127.0.0.1]) by tadpole.intranet (8.12.11/8.12.11) with ESMTP id i6HJxXWS065976; Sat, 17 Jul 2004 15:59:33 -0400 (EDT) (envelope-from lxv@tadpole.intranet) Received: (from lxv@localhost) by tadpole.intranet (8.12.11/8.12.11/Submit) id i6HJxWcc065960; Sat, 17 Jul 2004 15:59:32 -0400 (EDT) (envelope-from lxv) Date: Sat, 17 Jul 2004 15:59:31 -0400 From: Alex Vasylenko To: Julian Elischer Message-ID: <20040717195929.GA26144@tadpole.intranet> References: <40F835D0.8070609@elischer.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <40F835D0.8070609@elischer.org> User-Agent: Mutt/1.5.6i X-Scanned-By: MIMEDefang 2.43 cc: freebsd-current@freebsd.org Subject: Re: Call for PRs: nullfs X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list Reply-To: Alex Vasylenko List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 17 Jul 2004 19:59:37 -0000 On Fri, Jul 16, 2004 at 01:08:48PM -0700, Julian Elischer wrote: > I'm using nullfs for jails on -4.8(ish) without problems so far.. I find the performance of nullfs somewhat lacking as measured in the test described below (a config with nullfs performs worse (~2x slower) than the same config with vnodefs). For simplicity the test was done in chroot, doing it in a jail has no significant impact on performance. Hardware/OS: ------------ FreeBSD 5.2.1-RELEASE-p4 #1: Fri Apr 9 09:09:06 PDT 2004 CPU: Intel Pentium III (701.59-MHz 686-class CPU) Origin = "GenuineIntel" Id = 0x683 Stepping = 3 Features=0x387f9ff real memory = 268173312 (255 MB) atapci0: port 0xffa0-0xffaf at device 7.1 on pci0 ad0: 38166MB [77545/16/63] at ata0-master UDMA33 Config common: -------------- /var/tmp/world: populated with 'cd /usr/src; make installworld DESTDIR=/var/tmp/world' /var/tmp/jail: bin@ -> ro/bin dev/ etc/ lib@ -> ro/lib libexec@ -> ro/libexec ro/ root/ sbin@ -> ro/sbin tmp/ user/ usr/ var/ /var/tmp/jail/usr: X11R6/ bin@ -> ../ro/usr/bin include@ -> ../ro/usr/include lib@ -> ../ro/usr/lib libexec@ -> ../ro/usr/libexec local/ ports/ sbin@ -> ../ro/usr/sbin share@ -> ../ro/usr/share Config vnodefs: --------------- $ mdconfig -l -u md2 md2 vnode 131072 KBytes $ mount | grep /var/tmp/jail devfs on /var/tmp/jail/dev (devfs, local) /dev/md2 on /var/tmp/jail/ro (ufs, local, read-only) /usr/ports on /var/tmp/jail/usr/ports (nullfs, local, read-only) Config nullfs: -------------- $ mount | grep /var/tmp/jail devfs on /var/tmp/jail/dev (devfs, local) /var/tmp/world on /var/tmp/jail/ro (nullfs, local, read-only) /usr/ports on /var/tmp/jail/usr/ports (nullfs, local, read-only) Test: ----- chroot /var/tmp/jail cd /usr/ports/lang/perl5; time make WRKDIRPREFIX=/tmp Numbers: -------- nullfs: 176.257u 38.316s 6:04.13 58.9% 2801+2353k 183+366io 76805pf+0w vnodefs: 173.918u 33.775s 3:36.65 95.8% 2845+2407k 303+339io 251pf+0w base(*): 174.084u 33.281s 3:37.30 95.4% 2843+2415k 329+358io 252pf+0w (*) This was a simple 'cd /usr/ports/lang/perl5; time make' -- Alex.