Date: Sun, 24 Jan 1999 07:06:17 -0800 (PST) From: "Jonathan M. Bresler" <jmb@FreeBSD.ORG> To: grog@lemis.com Cc: jm7996@devrycols.edu, kheuer@gwdu60.gwdg.de, freebsd-advocacy@FreeBSD.ORG Subject: Re: FreeBSD - A User's Point of View Message-ID: <199901241506.HAA11804@hub.freebsd.org> In-Reply-To: <19990124201556.E36690@freebie.lemis.com> (message from Greg Lehey on Sun, 24 Jan 1999 20:15:56 %2B1030) References: <19990124171121.A36690@freebie.lemis.com> <Pine.BSF.4.05.9901240354030.215-100000@insomnia.local.net> <19990124201556.E36690@freebie.lemis.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> Date: Sun, 24 Jan 1999 20:15:56 +1030 > From: Greg Lehey <grog@lemis.com> > > > > The Linux filesystem, or ext2fs, if I'm not mistaken by default caches > > writes to the disk. If the machine should suddenly go down, power > > failure, unexpected crash, etc..., this information doesn't make it back > > to the disk. I've known many a Linux user who has lost _entire_ file > > systems due to this. > > UFS does this too. > [snip] > The only ``file system'' I know which doesn't cache significantly is > Microsoft's DOS file system. The performance is correspondingly bad. > > I think you've made my point. This ``evil'' feature of ext2fs was > probably borrowed, at least in concept, from BSD's UFS. FFS is significantly better than ext2fs. ext2fs does not order metadata writes. ext2fs does not use synchrouns writes for metadata. therefore the filesystems structures are not consistent until a shutdown or a long enough period of inactivity passes for the disk data to match that in cache. softupdates solves the problem by maintaining the dependencies on memory and keeping the filesystems on disk consistent. at least, thats my understanding. ;) jmb To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-advocacy" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901241506.HAA11804>