* Martin Webster [Tue, 16 Oct 2001 00:36:17 +0000]:
On Monday 15 October 2001 10:36 pm, Cliff Sarginson wrote: ( btw. I didn't write the following .. it was response to what I wrote) A journaling file system should be more robust since any changes are recorded in the journal before being written to disk.
Just don't forget that nearly all journaling file systems, at least by default, only journal meta data (directory and file info). There is nearly *no* data journaling, only ext3 offers it but at a speed penalty.
So journaling does save you time at fsck but it is no substitute for a regular backup.
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ? -- Regards Cliff
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ?
Faster FS checking is reason enough in itself, as anyone who's recently fscked a 40GB disk will tell you. That's mainly a server advantage. Plus you are immune (well, nearly) from FS corruption if the machine crashes. (But not data loss, as Phillip points out.) That's a server advantage too. Plus, for some workloads, particularly ones which involve many small files (like large project development) it's noticably faster. That's a workstation advantage. Don't feel obliged to change though. ext2 is still the choice of many and for good reason - mainly overall speed on generic workloads. -- 9:58am up 35 days, 23:11, 2 users, load average: 0.00, 0.00, 0.00
On Tuesday 16 October 2001 10.48, Cliff Sarginson wrote:
* Martin Webster [Tue, 16 Oct 2001 00:36:17 +0000]:
On Monday 15 October 2001 10:36 pm, Cliff Sarginson wrote:
( btw. I didn't write the following .. it was response to what I wrote)
A journaling file system should be more robust since any changes are
recorded
in the journal before being written to disk.
Just don't forget that nearly all journaling file systems, at least by default, only journal meta data (directory and file info). There is nearly *no* data journaling, only ext3 offers it but at a speed penalty.
So journaling does save you time at fsck but it is no substitute for a regular backup.
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ?
Compared to what? If you compare it to other journalling file systems, the question is up for grabs, as I don't know of any comparisons against JFS, XFS and ext3 having been done. Against ext2 the answer is 1) No corrupted file system. In case of crash you lose the data in the buffers that haven't been flushed yet, but the file system will never be corrupted. 2) It is significantly faster on small files, because of its revolutionary B-Tree inspired method of storing data. It is, to my knowledge, slightly slower at large files, but I don't think the difference is large enough to be called significant. 3) It is incredibly more efficient at storing small files. In ext2, a file always uses an integral number of blocks. Reiser stores files differently. ext2 with 4k block size: 2 files of 1k consumes 8k, in reiser they consume 2k. Yes, there is some overhead on metadata and journals, but believe me, I notice the difference on a server with many files. Ths more files, the bigger the difference. There was a problem in 2.4.0 with reiser, but since then AFAIK it's safe to use on servers. regards Anders -- ICQ: 132715168 YM: cicerose
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ?
Against ext2 the answer is
1) No corrupted file system. In case of crash you lose the data in the buffers that haven't been flushed yet, but the file system will never be corrupted.
2) It is significantly faster on small files, because of its revolutionary B-Tree inspired method of storing data. It is, to my knowledge, slightly slower at large files, but I don't think the difference is large enough to be called significant.
3) It is incredibly more efficient at storing small files. In ext2, a file always uses an integral number of blocks. Reiser stores files differently. ext2 with 4k block size: 2 files of 1k consumes 8k, in reiser they consume 2k. Yes, there is some overhead on metadata and journals, but believe me, I notice the difference on a server with many files. Ths more files, the bigger the difference.
There was a problem in 2.4.0 with reiser, but since then AFAIK it's safe to use on servers.
I have a little problem with reiser fs though. It seems it corrupts the "wtmp" file, at least in my systems. I got this problem only in my reiser fs machine, and I have noticed that someone else has got the same experience... I dont remember who and where. Praise
On Tue, 16 Oct 2001 11:36:25 +0200
Anders Johansson
On Tuesday 16 October 2001 10.48, Cliff Sarginson wrote:
* Martin Webster [Tue, 16 Oct 2001 00:36:17 +0000]:
On Monday 15 October 2001 10:36 pm, Cliff Sarginson wrote:
( btw. I didn't write the following .. it was response to what I wrote)
A journaling file system should be more robust since any changes are
recorded
in the journal before being written to disk.
Just don't forget that nearly all journaling file systems, at least by default, only journal meta data (directory and file info). There is nearly *no* data journaling, only ext3 offers it but at a speed penalty.
So journaling does save you time at fsck but it is no substitute for a regular backup.
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ?
Compared to what? If you compare it to other journalling file systems, the question is up for grabs, as I don't know of any comparisons against JFS, XFS and ext3 having been done.
Against ext2 the answer is
1) No corrupted file system. In case of crash you lose the data in the buffers that haven't been flushed yet, but the file system will never be corrupted.
2) It is significantly faster on small files, because of its revolutionary B-Tree inspired method of storing data. It is, to my knowledge, slightly slower at large files, but I don't think the difference is large enough to be called significant.
3) It is incredibly more efficient at storing small files. In ext2, a file always uses an integral number of blocks. Reiser stores files differently. ext2 with 4k block size: 2 files of 1k consumes 8k, in reiser they consume 2k. Yes, there is some overhead on metadata and journals, but believe me, I notice the difference on a server with many files. Ths more files, the bigger the difference.
There was a problem in 2.4.0 with reiser, but since then AFAIK it's safe to use on servers.
I've heard of serious problems with 2.4.11 and reiser fs but i don't rememberif it was on this ML ...
regards Anders -- ICQ: 132715168 YM: cicerose
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
On 16 Oct 2001, Emmanuel ESCARABAJAL wrote:
I've heard of serious problems with 2.4.11 and reiser fs but i don't rememberif it was on this ML ...
Hmm... I may be the person you are referring to... maybe. I believe the VM was "fixed" by 2.4.11 == 2.4.12. 2.4.10 corrupted my reiserfs filesystem *at least* 3 times (I didn't feel like using 'reiserfsck'... I just reformatted). That problem was reiserfs per se, it's just that the VM sucked _a lot_ at that point, especially under heavy load. Burning a CD at 12X and trying to run SETI@Home would cause the hard drive to thrash forever, at which point I would have to manually reboot and the partition was left corrupted. All this is on my 233mHz PI MMX with 32MB of RAM. BTW, 2.4.12 is nice and stable, at least for me. Now I can safely burn a CD at 12X and run SETI@home. -- noodlez: Karol Pietrzak PGP KeyID: 0x3A1446A0
On Tuesday 16 October 2001 17.19, Emmanuel ESCARABAJAL wrote:
I've heard of serious problems with 2.4.11 and reiser fs but i don't rememberif it was on this ML ...
I didn't know that. But then, I have heard plenty of other reasons not to use 2.4.11, so... :) Do you know if it's fixed in 2.4.12? regards Anders -- ICQ: 132715168 YM: cicerose
On Wed, Oct 17, 2001 at 01:26:48AM +0200, Anders Johansson wrote:
On Tuesday 16 October 2001 17.19, Emmanuel ESCARABAJAL wrote:
I've heard of serious problems with 2.4.11 and reiser fs but i don't rememberif it was on this ML ...
2.4.11 was a disaster and lasted approximately 2 days before LT put a fix in and released 2.4.12. The disaster was mostly apparent in Suse systems. All 2.4.11 kernels are marked as DONTUSE if you look around. -- Regards Cliff
It reportedly handles lots of little files better than ext2. I.e.,
good for a developer's workstation. See www.namesys.com for
details/propaganda.
HTH,
Jeffrey
Quoting Cliff Sarginson
* Martin Webster [Tue, 16 Oct 2001 00:36:17 +0000]:
On Monday 15 October 2001 10:36 pm, Cliff Sarginson wrote: ( btw. I didn't write the following .. it was response to what I wrote) A journaling file system should be more robust since any changes are recorded in the journal before being written to disk.
Just don't forget that nearly all journaling file systems, at least by default, only journal meta data (directory and file info). There is nearly *no* data journaling, only ext3 offers it but at a speed penalty.
So journaling does save you time at fsck but it is no substitute for a regular backup.
I think what I have yet to see an answer to is, aside from faster fs checking, what advantages does reiserfs confer. Or more precisely, a)What advantage does it give on a user workstation ? b)What advantage does it give on a server ?
-- I don't do Windows and I don't come to work before nine. -- Johnny Paycheck
participants (7)
-
Anders Johansson -
Cliff Sarginson -
Derek Fountain -
Emmanuel ESCARABAJAL -
Jeffrey Taylor -
Karol Pietrzak -
Praise