Date: Fri, 21 Sep 2001 12:09:36 -0700 From: Cy Schubert - ITSD Open Systems Group <Cy.Schubert@uumail.gov.bc.ca> To: j mckitrick <jcm@FreeBSD-uk.eu.org> Cc: Matt Dillon <dillon@earth.backplane.com>, Mike Meyer <mwm@mired.org>, "Daniel O'Connor" <doconnor@gsoft.com.au>, Kevin Oberman <oberman@es.net>, stable@FreeBSD.ORG Subject: Re: hw.ata.wc && hw.ata.tags && softupdates short question Message-ID: <200109211909.f8LJ9jH16272@cwsys.cwsent.com> In-Reply-To: Your message of "Fri, 21 Sep 2001 18:54:11 BST." <20010921185411.C81649@dogma.freebsd-uk.eu.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <20010921185411.C81649@dogma.freebsd-uk.eu.org>, j mckitrick writes: > > | You can choose one or the other, neither, or both. I typically run my > | laptop with softupdates turned on and write caching also turned on. A > | somewhat safer configuration would be softupdates turned off and write > | caching turned on, but not because softupdates and write caching might > | conflict. They don't. Only because softupdates itself delays meta-dat > a > | I/O and more work can be lost if a crash occurs then otherwise. > > Without knowing the details, I was just trying to sort out the tradeoffs of > each option. We've been told softupdates is reliable and makes UFS about as > fast as Linux ext2fs. We've been told write caching turned OFF causes a > huge performance hit, but ON can cause data loss. And we've also heard that > some controllers tell you WC is off even when it is still on. When you put > it all together, it makes quite an interesting fuzzy-logic truth table. :-) The reason that EXT2FS on Linux and UFS on Solaris and Tru64-UNIX are so fast is because by default those filesystems are mounted async. UFS on FreeBSD is mounted by default with SMD (Symmetric Meta Data), in that data is written asynchronously and metadata is written synchronously. SMD is more reliable than async but less reliable than synchronously mounted filesystems where the data is also written out synchronously. Softupdates orders writes of metadata such that you get almost the same speed (3% slower) as asynchronously mounted filesystems with the same benefits as SMD, except for the 30-60 second write delay. When the system crashes, there is greater risk of data loss over SMD or synchronously mounted filesystems. If you want almost absolute reliability, as absolute as it can get in this business, mount your filesystems synchronously and turn of write caching. If you want raw speed mount your filesystems asynchronously and turn on write caching. If we want even more speed, mount your filesystems with noatime. I personally have write caching turned on and use softupdates. It is a good compromise for what I need to do, which is a few desktops, firewalls, web servers, console and Kerberos servers (e.g. no database servers). It is also safer than what Linux, Solaris, and Tru64-UNIX do, which is by default asynchronously mounting their UFS and EXT2FS filesystems. Regards, Phone: (250)387-8437 Cy Schubert Fax: (250)387-5766 Team Leader, Sun/Alpha Team Internet: Cy.Schubert@osg.gov.bc.ca Open Systems Group, ITSD Ministry of Management Services Province of BC To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200109211909.f8LJ9jH16272>