From owner-freebsd-hackers Wed May 29 13:33:12 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id NAA03807 for hackers-outgoing; Wed, 29 May 1996 13:33:12 -0700 (PDT) Received: from Root.COM (implode.Root.COM [198.145.90.17]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id NAA03799 for ; Wed, 29 May 1996 13:33:09 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by Root.COM (8.7.5/8.6.5) with SMTP id NAA02323; Wed, 29 May 1996 13:32:29 -0700 (PDT) Message-Id: <199605292032.NAA02323@Root.COM> X-Authentication-Warning: implode.Root.COM: Host localhost [127.0.0.1] didn't use HELO protocol To: Terry Lambert cc: rashid@rk.ios.com (Rashid Karimov), davidg@Root.COM, jgreco@solaria.sol.net, hackers@FreeBSD.ORG Subject: Re: Breaking ffs - speed enhancement? In-reply-to: Your message of "Wed, 29 May 1996 13:24:37 PDT." <199605292024.NAA14236@phaeton.artisoft.com> From: David Greenman Reply-To: davidg@Root.COM Date: Wed, 29 May 1996 13:32:29 -0700 Sender: owner-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >> > >Am I just totally whacked out, or is this perhaps a reasonable thing to do, >> > >given that I'd really rather not have to absorb the extra write activity on >> > >the filesystems... does anybody else perceive any value along these lines >> > >of thought? >> > >> > I added an option "noatime" to mount/fstab and implemented a special >> > per-mount flag for this in the kernel. I was only interested in disabling >> > the access time; I wanted the inode change time and modify times to still >> > work correctly. My application, of course, was wcarchive - a machine with >> > millions of files that spends about 1/3-1/2 of all of it's disk I/O just >> > updating the access times in the inodes. >> >> David, could you submit those patches ? May be we can even add >> this as an OPTION to the kernel config file? > >There is a school of thought that says "shall be updated" in POSIX is >not the same as "shall be committed to stable storage" (the traditional >BSD implementation). > >This would let access times be updated in core, but only scheduled to >be written at a later time (not forced out immediately). They already are updated in-core and only written out during sync. The problem is that on busy machines, *thousands* of inodes have to be written out during the sync, and this can take 10+ seconds. With sync occuring every 30 seconds, this means the machine spends 33% of it's disk I/O time *just* writing out inodes. The access time is almost completely useless on a busy fileserver, so this is just a waste. -DG David Greenman Core-team/Principal Architect, The FreeBSD Project