Re: ext2 vs ext3 vs xfs vs reiserfs
Oleg said:
> Hi
>
> Can anyone who is very familiar with these FS types summarize their
> relative features in terms of
>
> a) quality (bug content)
> b) reliability (resistence to HD failures and system crashes)
> c) speed for workstation use
> d) compatibility (is it possible to convert from one FS type to another)
I've only worked with reiserfs on 2.2.x, I do not have any production
systems on 2.4.x yet, so xfs and ext3 are not options for me.
drawbacks to reiserfs:
- it does not handle bad blocks, so if your disks are of questionable
quality you may have trouble if they start mapping bad blocks in the area
where data is stored. e.g. 3ware raid controllers(6xxx series) do a piss
poor job of dealing with failed disks. Everytime a disk fails in on of my
arrays(3 systems with 6 disk raid10) the controller barfs, and reiserfs
panics the kernel saying there are bad blocks. Which should not happen because
the controller is supposed to mask the disk failure from the OS. There is
still another perfectly intact copy of the data on the 2nd disk.
- it does not support software raid on linux 2.2.x other then raid0, anything
else will cause data curroption. for this reason I stick to ext2 on my
software raid 1 systems.
- it constantly writes to the disk, so for laptop users the disk hardly ever
spins down
benefits:
- active development. I emailed hans about 8 months ago about a critical
bug in reiserfs that I experienced, within 24 hours I had a working fix.[1]
- reliable. I can't remember ever losing data to reiserfs, and I have
suffered at least I would guess 50-60 crashes accross multiple systems over
the past year.
- pretty fast. I only have ext2 to compare against but reiser is pretty
fast. filesystem performance isn't something I consider noticable though(by
the same token I don't notice any change going from p3-800 to athlon 1300
for most tasks)
- good quality. reiserfs has been under development for several years, much
longer then ext3 or xfs(linux port) from what I've seen. I've been reading
kernel traffic ever since it came out I think, and reiser has been around
the longest. To me, combined with the fact it's active development that
makes it the most mature. XFS has outstanding quality on IRIX, and an
outstanding track record on IRIX. It's still a baby in linux though, I am
still weary of it. ext3 sounds decent, but I prefer the ground-up design of
a journalling filesystem rather then tagging on to an existing one. that
said, ext3 has the only migration path that I'm aware of. Migration isn't an
issue for me, I'd happily wipe out my drives and reformat if needed.
keep in mind my reiser experience is limited to their 2.2.x branch of code,
I have not used it in a serious enviornment under 2.4.x Though I do have 1
SuSE system here which has it, along with a redhat 7.3 system that has
ext3, neither I consider critical to my home network.
I also appreciate that reiser realizes there are a lot of 2.2.x kernel
users out there and they continue to support them. EXT3 under 2.2.x was
abandoned more then a year ago last I heard, xfs has never worked under
2.2.x as far as I know, nor has JFS(IBM's fs). It's not perfect, but it
works well for me. And I don't need new features, I don't need reiser4
ported to 2.2.x, I just want working patches for the existing reiser for
whatever 2.2.x kernel is 'stable'. Currently my systems run 2.2.19.
nate
[1] in aonther thread a few months ago someone pointed out to me why
should I care if the filesystem is under active development. Well now
I point out for this specific reason. I encountered a very obscure reiserfs
bug which was ONLY triggered on 3ware raid 5 arrays(not raid 10 nor raid1),
the system would immedately reboot without warning upon triggering this
bug. It took a full day of work to track it down. But I recieved a fix
fast. If I encountered such a situation on ext3 for 2.2.x then I would
be SOL for the most part.
Reply to: