From: Sam Vilain (sam_at_vilain.net)
Date: Wed 12 Feb 2003 - 02:47:30 GMT
On Tue, 11 Feb 2003 02:29, Herbert Poetzl wrote:
> > In addition to data corruption, it's not all that hard to create a
> > directory structure that even root cannot read; I've just managed to
> > create one, and all I was doing was duplicating ~25% of the directory
> > structure using an analogue of `cp -al'. Reiserfs really cracks under
> > pressure, and that's the last thing you want a filesystem to do!
> > With these problems under high load, it's hard to think of a truly
> > useful application for reiserfs. It really is still experimental as
> > hell; the version in 2.4.20 seems particularly bad. Best to stick
> > with ext3/ext2
> Hans Reiser will hate you for that *G* ...
Well, if he hates me for that then he should focus on releasing code with
fewer bugs :-)
> > xfs/jfs if you really need the speed.
> seems you have detailed information about the
> speed/load issues compared between xfs, jfs, reiser
> and ext3? please share with us!
Oh, sorry, did I come across as an oracle? I'm actually full
of shit.
All I know about filesystem performance is;
1. getting a filesystem to work 100% is a long and drawn out process, and
requires intensive and widespread testing to iron out the bugs and
corner cases.
2. fundamental algorithmic enhancements, while seductively improving
the general case performance, increase and prolong the effect of
point 1.
3. stability cannot be measured with any benchmark or test suite.
I have also a basic and probably mildly inaccurate grasp of the algorithms
used in ext2 (two levels of bitmaps), xfs (b+tree(?)), jfs (b+tree) and
reiserfs (b*tree with gratuitous tail merging). I've read quite a few
benchmarks and comparisons, for each fitting the results in with my
understanding, and backed it up a little with some qualitative tests from
time to time.
When you consider how thoroughly debugged the code bases of jfs and xfs
are, their higher algorithmic complexity is complemented by loads of
debugging, and very switched on teams working with hardware that gives
them nice tidy dumps when it crashes for them to pore over, rather than a
hopeless `oops' message.
I don't care that much about it, really. All I want is a filesystem that
works. And reiserfs just doesn't bloody work properly.
-- Sam Vilain, sam_at_vilain.netReal software engineers don't write applications programs, they implement algorithms. If someone has an application that the algorithm might help with, that's nice. Don't ask them to write the user interface, though.