On Fri, 2010-05-21 at 03:36 +0000, yester64 wrote:
> Hi,
> not sure where to put this really, so i hope its ok here.
>
> I recently had my system running on reiser as my primary fs. Its not
> bad, but i noticed that it had some lag issues. Especially with database
> or copying from harddrive to harddrive.
> Accessing small files were quicker than with ext4 in my experience but
> the offset with the lag made it a poor choice for me.
>
> Are my finding valid, or is it just my system somehow?
Well… can’t say. Not sure what “lag” means. In general, we’ve been
pleased with reiserfs. We have probably over 50TB worth of reiserfs
filesystems.
What do we like? Reiserfs works VERY well with LVM and growing
filesystems is FAST and easy.
Ext3 (in SOME circumstances) can be grown online, but it can be VERY
slow due to the static nature of things in its design.
>
> Now i switched to ext3 since under ext4 some file went corrupt for some
> reason.
Ext4, while “blessed” (because it’s a Red Hat thing), hasn’t had a lot
of time to mature yet. There have been MANY fairly serious fixes for
ext4 even over the past year.
> I am happy for now and everything works like it should and no lag
> issue.
ext3 has some maturity… and works for most people. Just not quite as
flexible as reiserfs, but as you noted, some things are not well suited
for reiserfs (again, not sure what “lag” means).
>
> I recently read on the ubuntu forum that there are some plans to switch
> to btrfs, but that seems to be only a rumor.
> Are they any plans for suse? Or will it stay with ext4 for the future.
>
>
I think btrfs is the future… however, it has a WAY to go. Eventually,
I’m hoping we’ll be able to move away from reiserfs and onto btrfs. We
can’t really consider ext3/4, there’s just too much feature loss.
AFAIK, btrfs is being considered by multiple distros as the “future”
default filesystem, including openSUSE.
Ext4 is ok… just realize that it’s really just catching up to where
filesystem technology already WAS. There’s nothing revolutionary about
ext4. It’s just ext3 plus features that other filesystems have had for
6-8 years already.
So… will it be btrfs? Should we all use XFS? Maybe nilfs2? There
are lots of variables out there right now.
Do I think ext4 is the “future”… NO… absolutely, positively, NO.
Do I think that there will be even MORE “future” filesystems over the
next year or so besides ones mentioned here… YES, definitely yes.
IMHO, the next generation filesystem will handle LARGE scale filesystems
better, perhaps with object style replication across nodes, clustering,
etc. None of the available filesystems TODAY deal with large
filesystems (>2TB) effectively. I’m pretty sure that the winner(s) will
be the ones that can delete TBs worth of data in a second or two.
It may well be that log based filesystems like nilfs2 are the wave of
the future… just not sure if that one is the one that will make it.
IMHO, I like some of the ideas, but they need to be combined with some
other ideas with regards to handling of metadata (and others if
clustering, multi-node, etc are to be handled… exofs based
perhaps??).
IMHO, I’d like to see such a “golden” filesystem created with
distributed object based block handling and auto distributed object
mirroring and reconstruction with better/faster metadata handling… AND
have it called ninafs (for hopefully obvious reasons).
As far as lag goes , depends on what you are doing , databases, web, whatever. Many tuning options for the different filesystems out there.