Date: Sat, 7 Jul 2001 15:45:45 -0500 From: "Robert Banniza" <robert@rootprompt.net> To: <freebsd-stable@FreeBSD.ORG> Subject: RE: JFS Message-ID: <GMEDKMKMEBENJMBLDHAIMEIPEAAA.robert@rootprompt.net> In-Reply-To: <000d01c1074e$49d31ba0$0300a8c0@uhring.com>
next in thread | previous in thread | raw e-mail | index | archive | help
I have to agree with Dave on this. Working with large Sybase and Oracle installations, I see the advantages of transaction logging every day. While it is somewhat different than FS logging, I have wanted to see this on FreeBSD for some time. One of the reasons I keep Linux around rather than going with all FreeBSD is b/c of Reiser and friends. If logging had a downside in the industry, IBM, HP, SGI and SUN wouldn't be using it in their filesystems. Robert -----Original Message----- From: owner-freebsd-stable@FreeBSD.ORG [mailto:owner-freebsd-stable@FreeBSD.ORG]On Behalf Of Dave Uhring Sent: Saturday, July 07, 2001 8:36 PM To: Jim C. Nasby Cc: freebsd-stable@FreeBSD.ORG; Oliver Fromme Subject: Re: JFS ----- Original Message ----- From: "Jim C. Nasby" <jim@nasby.net> To: "Dave Uhring" <duhring@charter.net> Cc: <freebsd-stable@FreeBSD.ORG>; "Oliver Fromme" <olli@secnetix.de> Sent: Saturday, July 07, 2001 2:27 PM Subject: Re: JFS > That article does a good job describing RFS and contrasting it to > ext2fs, but it leaves several questions left un-answered in relation to > ufs. > > RFS might be faster or more stable than ufs, but there's no way to > assume that based on it's superiority to ext2fs, since ufs is also far > superior to ext2fs (I'm assuming ufs with softupdates turned on here). > > The ability to store data within the b-tree seems to be a valuable > feature. In fact, using a b-tree instead of other forms of metadata > storage is very interesting; it's similar to how most database engines > impliment indexes. > > I'm not so convinced that a journal is the greatest thing in the world, > though. I think it makes much more sense to do what softupdates do.... > perform updates to the filesystem atomically and in a manner where the > metadata is always written first. > > My $0.02, > dB! > > Dave Uhring wrote: > > > > On Saturday 07 July 2001 11:38, Oliver Fromme wrote: > > > A. L. Meyers <a.l.meyers@consult-meyers.com> wrote: > > > > Just installed SuSE Linux 7.2 with Reiser FS throughout on an > > > > Intel SMP box. The FS purrs, even on /, which doesn't mean > > > > everything is better or worse than FBSD. > > > > > > > > As far as I know, ReiserFS is GPL. What would porting it to > > > > FreeBSD be better or worse than other (newer and less stable) > > > > alternatives? > > > > > > Maybe a dumb question, but ... > > > > > > In what way is ReiserFS so much better than UFS + Soft- > > > updates + background fsck, that it would justify the > > > efforts of porting ReiseFS to FreeBSD? > > > > > > Just wondering. > > > > > > Regards > > > Oliver > > You seem to have missed the critical point of that paper. When the system goes completely haywire and either crashes or locks up so hard that a manual reset is required, UFS/softupdates requires a substantial amount of time to run fsck. If you have a very large filesystem, you then have to w....a....i....t until fsck completes. And if you are lucky, it will not terminate with the suggestion that you run fsck by hand. With a true journalling filesystem this wait is obviated. The last transactions are rerun or truncated and the system boots up. I use "logging" on Solaris and XFS on Linux and have tried reiserfs on Linux. All are superior to UFS/softupdates when the going gets tough. Disk access times may or may not be comparable with UFS/softupdates, but the integrity of my filesystems is more important than raw speed. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message 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?GMEDKMKMEBENJMBLDHAIMEIPEAAA.robert>