[NTLUG:Discuss] SuSE and the KDE folks (finally!) have their act together?

[Leroy Tennison]

Justin M. Forbes wrote:
> On Tue, Oct 06, 2009 at 10:44:46AM -0500, Chris Cox wrote:
>> ext4 is an "ok" filesystem, but NOTHING new.  New I suppose
>> for ext3 users, but ext4 doesn't present anything compelling.
>> Even recent benchmarks show crummy old reiserfs as being better
>> in many ways.  Ext4 needs more work... even so, then it will
>> be a ho-hum filesystem at best.  There's a reason why
>> XFS and reiserfs still reign in the datacenter.
>>
> ext4 is really supposed to be a small improvement over ext3.  The biggest
> improvements visible to users is not having to wait hours for a few TB of
> storage to be fscked.  
> 
>> I'd look to btrfs and nilfs2 to supplant ext4... and then,
>> I believe that there will have to be something even more
>> radically different coming in the next few years filesystem
>> wise.
>>
>> (begin slam)
>> But ext4 will be entrenched, but that's mainly because Red
>> Hat can't seem to deal with the idea of different filesystems.
>> (end slam)
> 
> Ext4 will still be widely used, but even in Red Hat we have people actively
> working on btrfs.  The problem is btrfs is nowhere near ready for
> production and still can cause data corruption.  Maybe by Fedora 14 it will
> be the default.  There was some initial discussion on ext4 vs btrfs for
> Fedora 11 and btrfs just wasn't ready.  In the meantime there are many
> users who are suffering from recovery times even on 1TB drives which are
> very common these days.
> Reisefs performs well, but has very little upstream effort to keep it
> stable and fix issues.  That certainly isn't good for enterprise customers
> at this point.  Even SuSE has switched from defaulting to resierfs.
> XFS can be a great filesystem for large volumes, but is not as flexible for
> smaller volumes.  It also seems to get a very large number of fixes where
> things have broken upstream. Luckily very few of those are data corruptors,
> but I wouldn't call it the pinnacle of stability either.
> In the end, performance is very important, but data integrity will always
> be more important.  There is no better way to lose customers than losing
> their data.  I think that everyone is looking for a performance improvement
> over ext3/4 but shipping something that isn't ready is just asking for
> pain.
> 
> Justin M. Forbes
> 
> _______________________________________________
> http://www.ntlug.org/mailman/listinfo/discuss
> 

If fsck is faster, how about searching for files?