[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ext3 on 2.4.4



Bryan Smith writes:
> I have heard that you have dropped all intensions to extend Ext3
> beyond the capabilities/features of Ext2/VFS.  If this is true, that
> means that Ext3 is a "temporary JFS solution" -- although don't read
> much into that because I defend Ext3 existance almost daily against
> those who I like to call "ReiserFS Absolutists."  In dealing with
> these "misguided" individuals, I find myself trying to break down
> all the capabilities, features, concerns, approaches (e.g., Ext3 =
> evolutionary, ReiserFS = revolutionary, XFS = existing ports) of the
> various JFS options for Linux.

Actually, even if we "stopped" development of new ext2/ext3 features
today and just merged in the existing features into the stock kernel,
ext2/ext3 would be a first-class citizen of filesystem features (and
hopefully also performance) for a long time to come.

Still-to-be-merged features include indexed directories (handling for
very large number of files in a single directory = millions of files).
Online resizing of the filesystem (in conjunction with LVM, et. al.).
Extended attributes/Access Control Lists for security, extra metadata.
Tail merging to handle lots of files smaller than the filesystem block
size.

One of the great benefits of ext2 is that it is designed from the start
to be extended in such ways.  Combined with the fact that e2fsck is
very robust at fixing filesystem errors (which is _much_ harder to do
on tree-based filesystems), that ext2 has been in use for a long time
to fix the corner cases, and that a large number of people are looking
at the ext2 sources to fix bugs and improve performance, I think ext2
will continue to be well supported for a long time to come.

> Again, I'd appreciate any insights you can provide.  I'm fairly
> active with several LUGs here in the Florida area and want to make
> sure I'm pumping out technically sound analysis and not opinionated
> technobable when talking about Ext3 and XFS admist "ReiserFS
> Absolutists."

Obviously I can't respond for Stephen (and I doubt he can/will respond
"as a RedHat employee" as opposed to his personal views).

Unfortunately, some ReiserFS benefits are based on self-serving
technobabble themselves(*), so you will have a hard time convincing
people with just facts...

(*) See Documentation/Configure.help:CONFIG_REISERFS_FS for an example.
    I doubt you will find another entry which talks about "future features",
    or solicits money for features and support.  It is also doesn't
    surprise me that their patch added ReiserFS at the very top of
    fs/Config.in, while other filesystems are generally alphabetical.

Also, while XFS is a feature-full filesystem I have read that the source
code is _just_huge_ (i.e. a large percentage of the full Linux kernel source
size) so while it has some performance and feature benefits, it also hits
you right in the cache/CPU.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]