On Thu, Dec 28, 2000 at 01:51:26PM -0500, Jon Lindgren wrote:
> I've been itching for a while for the ability to run a cluster of machines
> which share a common physical disk.
> 
> I've been thinking of starting to learn the filesystems code, and perhaps
> implement a simple [heh!] proof-of-concept type of f/s sharing, perhaps
> using a userland daemon to maintain locking, etc...  I'm not looking for
> tremeoundous performance or stability at this point; just a kind of "look,
> it works okay, neat-o" type of implementation.
> 
> Of course, this is gonna use a SCSI bus, probably maintaining locks over a
> network connection (seeing as the SCSI subsystem doesn't support talking
> adapter to adapter at this point).  I'd probably just hack up a version of

This may change soon :)

> ffs to use as the filesystem.  Again, I'm not looking for a fantastically
> superb implementation - just a "that was fun, look mom, two machines - one
> disk".
> 
> Is anybody already working on such a project?  Does anyone have any great
> ideas or pointers?  At the very least, I'd want to take the ffs code and
> copy it to a new f/s name (probably "hackfs" or such).  I'd also love any
> ideas on how to talk from the kernel to a userland daemon (never had to do
> such a thing).

I'd like to see something like this, but I don't have more idea than
that :)
 
--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--