From owner-freebsd-current Tue Jan 19 14:47:28 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id OAA19454 for freebsd-current-outgoing; Tue, 19 Jan 1999 14:47:28 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from smtp01.primenet.com (smtp01.primenet.com [206.165.6.131]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id OAA19447 for ; Tue, 19 Jan 1999 14:47:26 -0800 (PST) (envelope-from tlambert@usr04.primenet.com) Received: (from daemon@localhost) by smtp01.primenet.com (8.8.8/8.8.8) id PAA13496; Tue, 19 Jan 1999 15:47:14 -0700 (MST) Received: from usr04.primenet.com(206.165.6.204) via SMTP by smtp01.primenet.com, id smtpd013430; Tue Jan 19 15:47:10 1999 Received: (from tlambert@localhost) by usr04.primenet.com (8.8.5/8.8.5) id PAA05080; Tue, 19 Jan 1999 15:47:05 -0700 (MST) From: Terry Lambert Message-Id: <199901192247.PAA05080@usr04.primenet.com> Subject: Re: ufs async To: ck@adsu.bellsouth.com (Christian Kuhtz) Date: Tue, 19 Jan 1999 22:47:04 +0000 (GMT) Cc: freebsd-current@FreeBSD.ORG In-Reply-To: <19990117112925.A19618@oreo.adsu.bellsouth.com> from "Christian Kuhtz" at Jan 17, 99 11:29:25 am X-Mailer: ELM [version 2.4 PL25] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > Hey gang: > > Somebody mentioned the other day that UFS async isn't entirely async, and > that directories are (always) sync. Is that true? No. The default is sync metadata and async user data, with the assumption that the user will have taken a CS 203 class and know about two stage commit for database/index integrity, if they care about their data. With the "sync" option, all data, both metadata and user data, is written synchronously. With the "async" option, all data, both metadata and user data, is written asynchronously. Put another way: default: A seat belt for you, but not the passenger sync: A seat belt for you and your passenger async: You are both strapped to the bumper There's also soft updates. The point of soft updates is to ensure ordering for maximal data ingrity (default ensures only metadata integrity by ordering metadata operations, sync assures only that writes occur in order, but that user data/metadata dependencies are not assured because of implied dependecies between file contents, and async implies no data integrity whatsoever). Basically, soft updates is just a more efficient way of ensuring operations occur in order (like the USL "DOW" -- Delayed Ordered Writes -- technology, which is under patent, only more efficient). So... soft updates: A seat belat and an air bag for both you and your passenger Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message