From owner-freebsd-hackers Wed Feb 4 21:40:21 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id VAA03887 for hackers-outgoing; Wed, 4 Feb 1998 21:40:21 -0800 (PST) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from alpo.whistle.com (alpo.whistle.com [207.76.204.38]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id VAA03877 for ; Wed, 4 Feb 1998 21:40:10 -0800 (PST) (envelope-from julian@whistle.com) Received: (from daemon@localhost) by alpo.whistle.com (8.8.5/8.8.5) id VAA09931; Wed, 4 Feb 1998 21:38:27 -0800 (PST) Received: from UNKNOWN(), claiming to be "current1.whistle.com" via SMTP by alpo.whistle.com, id smtpd009928; Wed Feb 4 21:38:18 1998 Message-ID: <34D94F70.237C228A@whistle.com> Date: Wed, 04 Feb 1998 21:34:40 -0800 From: Julian Elischer Organization: Whistle Communications X-Mailer: Mozilla 3.0Gold (X11; I; FreeBSD 2.2.5-RELEASE i386) MIME-Version: 1.0 To: hackers@FreeBSD.ORG CC: mckusick@mckusick.com Subject: soft-updates performance differences. Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG X-To-Unsubscribe: mail to majordomo@FreeBSD.org "unsubscribe hackers" These numbers were found today running soft-updates on freebsd-current (as of 3 days ago).: The tests are simplistic and do not reflect any known meaningful work other than hierarchical copies and deletes. method: mount a 64MB soft-update filesystem as /mnt time cp -pr /usr/X11R6 /mnt time rm -rf /mnt/X11R6 ----------------------------- normal sync mount: COPY user time: 0.1 sec sytem time: 4.9 secs real time: 2:16 6339 sync writes, 3788 async writes RM user time 0 sec system time: 0.9 sec real time: 0:59 5884 sync writes, 74 async writes ------------------------------- normal mount -o async COPY user time: 0.1 sec system time: 4.6 secs real time: 1:08 325 sync writes, 3873 async writes RM user time: 0.0 system time: 0.8 secs real time: 0:26 secs 2941 sync writes, 163 async writes (why so many sync?) --------------------------------- soft updates mount (after tunefs -n enable) COPY user time: 0.1 secs system time: 5.1 secs real time: 1:13 secs 0 sync writes, 4916 async writes RM user time: 0 secs system time: 0.5 secs real time: 0:01.27 (!!) 0 sync writes, 238 async writes ------------------------------------- notes. The the free space for the RM on soft updates is recovered over the 30 seconds following the rm. It's amusing doing a 'df' and seeing it come back even though you are not doing anything. A system failure at any time during the soft-updates mount would leave a disk that is completely consistant, except for the fact that some blocks may not be freed (wow) unlike the -o async case. Note especially that 'consistant' does not mean 'no data loss'. Data written to disk in the last 30 seconds and NOT FOLLOWED BY A FSYNC may not be present after reboot... If you plan on hitting 'reset', you need 3 syncs to ensure everything you hope will be on the disk actually is.. (at this stage) (you shouldn't be doing this anyhow!) Soft-updates is a step forward in the technology but is not a magic bullet.