how to convert to ReiserFS ? (Answer: Don't - Reiser isn't that great)

[Gabriel M Dragffy]

On Fri, 2006-07-28 at 15:40 -0400, Nate wrote:

> I've had a similar bad experience with ReiserFS. I've gone for the whole 
> Reiser thing twice. Each time, the filesystems eventually became 
> corrupt, and before they did that, they were always working the disks 
> and sucking up memory. ReiserFS, alhtough fast for some things, sucks up 
> an /incredible/ amount of memory. Reiser4 actually takes up even more. 
> Even with a 512 MB laptop, this was a little much. I'm now happy with 
> JFS, which is on average faster than Reiser, more stable, and uses less 
> memory. The only drawback is that boot-time takes about three-four 
> seconds more because of the occasional automatic fsck.jfs.
> 
> Want me to back up my comments about JFS being faster and taking up less 
> memory than both the Reisers? Why, sure.
> 
> *(Aug, 2001: Theurer does a benchmark test: JFS has much greater data 
> throughput than Reiser. http://lwn.net/2001/0830/a/jfs-comparison.php3)
> 
> *(Nov, 2001: Scalzo shows that Reiser is shown to be terrible, terrible, 
> terrible with RAID and LVM systems. 
> http://www.quest-pipelines.com/newsletter-v2/linux2.htm)
> 
> *(Oct, 2003: Benoit does extensive benchmarking, coming up with the 
> result that for everyday computing, CPU limited applications, and bang 
> for your buck, JFS is the way to go, the last field there accompanied by 
> XFS. Reiser is shown to have some potential advantage in heavy, 
> operations distributed over hundreds of small files, and on SCSI disk 
> drives (which the average user doesn't have), but at the cost of a vast 
> amount of memory.  JFS is shown to perform an incredibly high ratio of 
> Work:CPU usage. Also, JFS is shown to take less time to read and write 
> large files [although XFS is the unquestionably the ultimate FS in 
> managing large files, JFS compares very favorably to it in these tests 
> when you look at the metrics.])
> 
> *(May, 2004: Piszcz's first benchmark test, a very famous one: ReseirFS 
> has some specific uses that it might be better for, but JFS is about the 
> same speed, and takes up a whole lot less memory - 
> http://linuxgazette.net/102/piszcz.html)
> 
> *(January, 2006: Piszcz's second article, in which he goes through much 
> more meticulous and extensive tests, and this time with the more recent 
> 2.6 linux kernel. JFS proves to be much faster than ReiserFS and 
> Reiser4. http://linuxgazette.net/122/TWDT.html#piszcz)
> 
> *(April 2006: Sivers performs a home-server centered benchmark test and 
> releases the results on an article on Debian-Administration.org 
> Apparently: XFS is great for server, JFS is great for small-time 
> applications. ResierFS is STILL very buggy [after all these years of 
> supposed active 'bug-fixing.' I don't want to hear any more junk about 
> how "they've pretty much fixed all the bugs"] - 
> http://www.debian-administration.org/articles/388)
> 
> *(July 2006: This page has an interesting discussion about the 
> misleading and selective "benchmarks" on the namesys site 
> http://osnews.com/read_thread.php?news_id=15129&comment_id=141566)
> 
> Those benchmarks are the same ones which only show Reiser vs. Ext3 (just 
> about one of the slowest FSs you can get.) These are the same benchmarks 
> that only show results in which reiser does favorably, the ones which he 
> performed with "write" and "append" mode off (Reiser can only go that 
> fast when it doesn't include file-editing capability, a change would 
> speed any FS up dramatically), the ones which he was called on for being 
> misleading two years ago, and he said he was wrong and "forgot about it" 
> but to this date has not changed those benchmark pages. Not to match up 
> with new, up to 30% faster versions. Not to fill in the selectively 
> neglected benchmarks. Nothing. What a nice, considerate, honest guy, 
> huh? Now why would he need to do something like that if his FS was as 
> good as he said it was? The answer is: it isn't as good as he says it 
> is, as just about any major benchmark will show.  It's all a bunch of 
> hokum pokum and lies.
> 
> Face it, guys, namesys (the people who make reiserfs under the 
> supervision of Hans himself.) has been pulling the wool over your eyes 
> about their "constant improvement and perfection". They're a whole bunch 
> of marketers trying to make money off their shoddy FS creation by 
> transitioning large businesses to their filesystem, and in a field that 
> is relatively virgin to corporate scheming, linux users generally have a 
> hard time realizing this. They're a company, Reiser is Han's baby. They 
> will lie and do anything they can to make the FS seem effective, because 
> pride and profit are on the line and that's what they value most. They 
> do alright work, but they lie about it and spin it all over and hype up 
> the naive to the degree that they appear revolutionary. The thing is, 
> it's just another FS, and not a very good one at that. Better than ext3, 
> but uses more memory, and is significantly worse for the home computer 
> and home server than JFS and XFS.
> 
> This is linux, so if you'd like to use reiser, feel free to do so. If 
> you're using reiser happily, by all means continue to do so. However, if 
> you're looking for a fast, light, and journaled file system, I'd suggest 
> you look into JFS instead of Reiser. For one thing, JFS is generally 
> faster and takes up less memory, for the other, Reiser causes a lot of 
> trouble. On two different distributions, when I used reiser (and I never 
> exceeding about 20-40% of partition use), the partition eventually 
> became corrupted. There was nothing I could do about it, and any 
> attempts I made at recovery were nonfunctional (when I finally found a 
> livedisk that did support reiserfsck, it didn't help, the partition was 
> just broken beyond repair). If JFS ever fails you, with JFS being almost 
> an industry standard (figure of speech), almost every livedisk has 
> fsck.jfs, and that tool has never failed me.
> 
> You thinking about switching filesystems? For stability, speed, and 
> conservation of system resources, I'm going to preach what has worked 
> incredibly for me and ask you to consider JFS in your choice. And I'm 
> going to suggest that if you do go for ReiserFS, maintain your skepticism.


Amazing! To be frank I didn't believe all your criticisms of reiser, I
have been using it for about two years and even after power failure it
has always recovered very gracefully and quickly. Anyway I've been
meaning to switch to LVM for a while and just today I did that, taking
this opportunity to try out JFS.
My opinion.... unbelievable! My system doesn't take 3-4 seconds longer
to boot up, it is noticeably quicker, I'd say half the time. That's with
minimal exaggeration. My hard drive is so quiet i don't even hear it
scraping around. Ubuntu is so much more responsive, even clicking on the
menus such as SYSTEM/PLACES/APPLICATIONS show up immediately with all
the icons, instead of taking half a second to display the icons. The
only thing I found it was slower at is deleting lots of files, but JFS
hardly takes any time. I also did my own benchmarking of copying around
a 1.2GB file, reiserfs kicked JFS's ass, so maybe the way to go is JFS
on / and Reiserfs on /home. However I am going to experiment with using
a pure JFS system for a while and see how it fares generally.

Just in case anyone was wondering what system I'm running on: it's a
sony vaio laptop which i bought new, 1 year ago, so it's pretty good. It
seems that the component are obviously made with efficiency and not
performance in mind and my guess is that the resource-intensive reiserfs
chokes my laptop.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-users/attachments/20060730/4a4b113d/attachment.html>