Hi!
I beg to differ.
If high performance appeals to you, give ReiserFS OR XFS a try - since
all benchmarks on this subjects show a more or less significant
difference dealing with a large number of big vs. small files.
I also don't quite understand the suggestion to ignore "arguments about
data corruption". These weren't arguments but simple facts. A lot of
posters here experienced various troubles with almost every FS there is.
That doesn't proof that any of the discussed filesystems is BAD - but it
does proof that - excuse my english - shit happens and you might lose
data regardless of the FS you went for. A 100% bulletproof FS simply
seem not to exist.
When I had to decide the FS issue myself my questions hovered over
things like
- performance / benchmarks
- journaling quality
- recovery tools
- compatibility (parted issues, non-linux OS drivers and so forth)
- FS overhead
- etc. etc.
Regards
spox
Am Dienstag, den 19.04.2005, 08:16 +0200 schrieb Richard Fish:
> Jarry wrote:
>
> > I'm counting votes, and waiting for some final decision to come.
> > I can not contribute to this discussion, because I have absolutely
> > no experience with journaling filesystems at all. That's why I
> > asked...
> >
> > Up to now I'm more confused than before posting my question.
> > Anyway thanks to all who replied.
> >
>
> I would summarize it this way:
>
> If high performance appeals to you, give reiserfs a try.
>
> If full data journaling (instead of just meta-data) appeals to you, go
> with ext3, but read up on the tuning options available. I should note
> that for some very specific workloads, full data journaling with ext3
> will be faster than reiserfs.
>
> I would ignore the arguments about "x filesystem corrupted my data, and
> y never has". Personally, I had filesystem corruption with reiserfs
> back in '99 or so. It interacted badly with the VM in a few of the 2.4
> kernel series. In another case, I had faulty hardware. But data
> corruption is not widespread for any of (xfs, reiserfs, ext3, jfs) with
> current and properly configured kernels and working (especially not
> overclocked) hardware.
>
> Pick one and use it.
>
> -Richard
>
--
Mit freundlichen Gr��en
Heinz Sporn
SPORN it-freelancing
Mobile: ++43 (0)699 / 127 827 07
Email: [EMAIL PROTECTED]
[EMAIL PROTECTED]
Snail: Steyrer Str. 20
A-4540 Bad Hall
Austria / Europe
--
[email protected] mailing list