From owner-freebsd-fs@FreeBSD.ORG Thu Feb 12 18:48:39 2004 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id DEE7B16A4CE for ; Thu, 12 Feb 2004 18:48:39 -0800 (PST) Received: from beelzebubba.sysabend.org (alcatraz.inna.net [209.201.74.14]) by mx1.FreeBSD.org (Postfix) with ESMTP id A66E243D1F for ; Thu, 12 Feb 2004 18:48:39 -0800 (PST) (envelope-from xyzzy@moo.sysabend.org) Received: from moo.sysabend.org (moo.sysabend.org [66.111.41.70]) by beelzebubba.sysabend.org (Postfix) with SMTP id 1C9F014E9F; Thu, 12 Feb 2004 21:48:54 -0500 (EST) Received: (nullmailer pid 83742 invoked by uid 14); Fri, 13 Feb 2004 02:48:38 -0000 Date: Thu, 12 Feb 2004 18:48:38 -0800 From: Tom Arnold To: Xin LI Message-ID: <20040213024837.GQ13780@moo.sysabend.org> References: <20040212235605.GP13780@moo.sysabend.org> <20040213022010.GA2331@frontfree.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20040213022010.GA2331@frontfree.net> User-Agent: Mutt/1.4i Organization: The Sysabend Dump X-Operating-System: CPM2.2 X-Bucket-Brigade-Devices: Rah! X-8-Bit-Samples-And-Analog-Filters: Rah! cc: freebsd-fs@freebsd.org Subject: Re: Upcoming filesystems? X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list Reply-To: xyzzy@sysabend.org List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 13 Feb 2004 02:48:40 -0000 On Fri, Feb 13, 2004 at 10:20:10AM +0800, Xin LI wrote: > Personally I'd prefer the SoftUpdates approach, so I am more interested > in what made you to think it doesn't fit your needs, performance? > security features? or others? Boot time. Large small-block filesystems take forever to come up at bootime Forever being approx 10 minutes per filesystem for the snapshot to occur, with 3 filesystems on the machine it adds up. If you go really insane on your block size on a large enough disk, in my case 1TB disk, 4k block size, 2billion inodes, the snapshot will just error out saying it doesnt have enough space. This isnt as much of an issue as we've determined the average filesize to be markedly over 512 bytes and it seems to handle 8k/1k just fine, but I can see this sneaking up on someone as drives get larger and larger... Minor issue, there are times when you cannot remove a file or subdirectory while the bgfsck is running. Important to us because of how the space is being used. Has to be large numbers of small blocks/fragments because its an extreme numbers of small files. I may not agree with how its being done, but I have to deploy a machine to handle it. I can expect hundreds of millions of 1k files on the each 1TB slice of a 3TB array. The boot-time is whats really killing us. Performance-wise, I've found softupdates to easily outperform ReiserFS. -- ------------------------------------------------------------------------ - Tom Arnold - When I was small, I was in love, - - Sysabend - In love with everything. - - CareTaker - And now there's only you... - -------------- -- Thomas Dolby, "Cloudburst At Shingle Street" -