opinion on XFS

[Donald Becker]

On Wed, 8 May 2002, Jarrod Smith wrote:

> On Tue, 7 May 2002, Donald Becker wrote:
>
> > The SGI release of "Linux XFS" was a port of the Linux code to XFS, not
> > a XFS implementation for Linux.  The port was done in a way that
..
> I can certainly understand how someone with a large investment in the
> inner workings of the Linux kernel would feel this way.  However, from a
> user/administrator standpoint XFS is a wonderful option to have.

I guess my comments came across as negative, rather than just factual.

I would love to have XFS.  Our cluster system is "filesystem agnostic"
-- we don't need any filesystem at all(*).  Any filesystem that can be
loaded as a module can be used to dynamically enhance the system, with
no impact on the cluster architecture.  The problem is that XFS uses a
different interface, and thus can't work with a standard kernel.

(*) We actually use a ram-based filesystem to boot the cluster nodes.
But that doesn't count ;->.

> CXFS on Linux will be nice if it sees the light of day, but from what I
> understand this will be a commercial product, not open source like XFS.
> In any case, I really hope XFS makes it into the next kernel.


-- 
Donald Becker				becker at scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993