From owner-freebsd-hackers Wed Nov 12 06:28:19 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id GAA06791 for hackers-outgoing; Wed, 12 Nov 1997 06:28:19 -0800 (PST) (envelope-from owner-freebsd-hackers) Received: from mail.virginia.edu (mail.Virginia.EDU [128.143.2.9]) by hub.freebsd.org (8.8.7/8.8.7) with SMTP id GAA06785 for ; Wed, 12 Nov 1997 06:28:13 -0800 (PST) (envelope-from atf3r@cs.virginia.edu) Received: from ares.cs.virginia.edu by mail.virginia.edu id aa22736; 12 Nov 97 9:28 EST Received: from stretch.cs.virginia.edu (atf3r@stretch-fo.cs.Virginia.EDU [128.143.136.14]) by ares.cs.Virginia.EDU (8.8.5/8.8.5) with ESMTP id JAA07782; Wed, 12 Nov 1997 09:28:06 -0500 (EST) Received: (from atf3r@localhost) by stretch.cs.virginia.edu (8.8.5/8.8.5) id JAA11610; Wed, 12 Nov 1997 09:28:05 -0500 (EST) Date: Wed, 12 Nov 1997 09:28:04 -0500 (EST) From: "Adrian T. Filipi-Martin" Reply-To: adrian@virginia.edu To: Tom cc: Alfred Perlstein , hackers@freebsd.org Subject: Re: LFS system? In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk On Tue, 11 Nov 1997, Tom wrote: > > On Wed, 12 Nov 1997, Alfred Perlstein wrote: > > > has any though been put towards a log based File system? > > Lots. See mount_lfs code (keep in mind that it doesn't work). Do a > "man -k lfs" to get the whole picture. Yep, lots. Ousterhout, one of the designers of LFS, was at Berkely. I belive the BS support was put in by his group originally. The concept was first implemented under the SPRITE OS. It just hasn't been maintained in the 4.4BSD's though. :-( > > would it be a performance gain at all? i think it could be a major > > improvement on heavily modified file systems for instance on a large News > > server were a sync might take a few seconds to complete. > > Performance gain? I always though LFS files systems were slow, due to > the extra overhead. However, that overhead buys you security, and fast > filesystems checks during start up. LFS is fast for filesystems with lots of write activity. LFS is optimized for large reads and writes by always writing tracks and such in a single pass. Futhermore, disk blocks are never updated, they are just rewitten to a new location. The clean-up you allude to does need to be done asynchronously though by a cleaner-process. It is akin to garbage collection which make alot of people wrongly assume it has bad runtime performance or predictability. The log structure does provide nive data security, but it is also a performance booster. For more info look for a paper called "Breaking the I/O Bottleneck" by Ousterhout, et al. Adrian -- adrian@virginia.edu ---->>>>| If I were stranded on a desert island, and System Administrator --->>>| I could only have one OS for my computer, Neurosurgical Visualzation Lab -->>| it would be FreeBSD. Think about it..... http://www.nvl.virginia.edu/ ->| http://www.freebsd.org/