I know there was a thread on this a while back, but now with reiserfs 4.0 to be released this December, how do these 3 journaling FSs compare and contrast. Any jotchas for each? Thanks, Babu __________________________________________________ Do You Yahoo!? HotJobs - Search Thousands of New Jobs http://www.hotjobs.com
Alle 01:57, giovedì 22 agosto 2002, babu walad ha scritto:
I know there was a thread on this a while back, but now with reiserfs 4.0 to be released this December, how do these 3 journaling FSs compare and contrast. Any jotchas for each?
Thanks,
Babu
I have been benchmarking them all with bonnie++ with the following results: reiserfs and xfs are much faster than jfs. jfs is just the slowest filesystem available on SuSE. Then here it is my thought: reiserfs also has some reliability problems and it is very slow in some cases (this is a rumor) even with small files, which should be the strenght of reiserfs instead. I would choose xfs if you are planning to remain with SuSE kernel, reiserfs if you are planning to use also vanilla kernels, as AFAIK xfs is not part of standard kernel. Note: both xfs and reiserfs were a bit faster than standard ext2 filesystem, which was a bit faster than ext3. ext3 is a journaling filesystem based on ext2 and it is the only one which make sure data coherence, not only filesystem coherence. Praise
Praise wrote: ...
Then here it is my thought: reiserfs also has some reliability problems and it is very slow in some cases
Could you please give examples, proof? I have never heard of any such problems, except for - in the very beginning, SuSE Linux 6.3, 6.4, etc. - at the beginning of switch to 2.4 kernel (which was a stability problem itself) - when people use some old(er) kernel, e.g. 2.4.7 (actually, just guessing, in fact I've not heard of any problems with reiserfs there) It's not really helpful when people just spread words they heard from a friend who read it somewhere on a mailinglist where someone else posted a message just like yours... ;-) I would not venture to speak for or against any of the fs's. And "bonnie" is aything but a real-world benchmark and I would not put too much emphasize on the results. I had serious fs trouble on my reiserfs based Sony Vaio notebook once, but I would definitely not want to attribute this to reiserfs itself - the notebook had lots of troubles like "suspend" would not always work and sometimes crash the system until I did a re-install of the Windows98 on it from the recovery disks, I haven't had *any* troubles since (for many months). So I'd be very careful before I blame anything! The real problem might be something completely different, computers are complex systems! Anecdotal evidence should only be permitted when it's really overwhelming, one bad experience doesn't show anything - see mine, from todays point of view it doesn't look like a reiserfs problem at all, but guess what I blamed it on when I lost ALL my files (and found out reiserfs does have a /lost+found/ ;-) )? Correct, reiserfs...
(this is a rumor) even with small files, which should be the strenght of
Same problem. Rumours are totaly useless. Regards, Michael
On Sat. Aug. 24, 2002 at 10:03:36 -0700 GMT, a lone cry was heard from
Michael Hasenstein
Could you please give examples, proof? I have never heard of any such problems, except for - in the very beginning, SuSE Linux 6.3, 6.4, etc. - at the beginning of switch to 2.4 kernel (which was a stability problem itself) - when people use some old(er) kernel, e.g. 2.4.7 (actually, just guessing, in fact I've not heard of any problems with reiserfs there)
Please refer to this article at IBM: http://www-106.ibm.com/developerworks/linux/library/l-fs7/ Charles -- Why use Windows, since there is a door? (By fachat@galileo.rhein-neckar.de, Andre Fachat)
Charles Philip Chan wrote:
On Sat. Aug. 24, 2002 at 10:03:36 -0700 GMT, a lone cry was heard from Michael Hasenstein
in the wasteland called the Internet: Could you please give examples, proof? I have never heard of any such problems, except for - in the very beginning, SuSE Linux 6.3, 6.4, etc. - at the beginning of switch to 2.4 kernel (which was a stability problem itself) - when people use some old(er) kernel, e.g. 2.4.7 (actually, just guessing, in fact I've not heard of any problems with reiserfs there)
Please refer to this article at IBM:
Thanks, very good resource, but the "stability" question is not answered. I just want to point that out, since the email that started this asked about just that. The author looked at what the fs's DO and how it's implemented, for the stability question one would have to find a way to reliably collect some good data from thousands of people... because even if it's stable on one own stress test system there's always someone out there with some really strange setup/workload/test case which no one thought of ;-) Michael
participants (4)
-
babu walad
-
Charles Philip Chan
-
Michael Hasenstein
-
Praise