On Sep 05 2003, at 10:00, der.hans was caught saying:
> Am 04. Sep, 2003 schw?tzte Shadow so:
> 
> > How about OpenAFS?  Unified directory space, transparent data migration,
> > multiple server support, expandable, LDAP/Kerberos authentication - the
> > only two things I would like to see is a Network based RAID and
> > disconnected operation ala Coda.
> 
> Look at the enhanced network block device.
> 
> Description: Enhanced Network Block Device support.
>  enbd-client proxies kernel requests, passing them across the net to a
>  remote enbd-server daemon and making a remote resource on the server
>  appear as a local block device.  NBDs may be treated like any other
>  block device.  In particular, they may have file systems made on them
>  and then may be mounted into the local file system tree.  This package
>  contains the userland support daemons for the Enhanced NBD kernel
>  driver.
>  .
>  Regular updates will be found at
>  ftp://oboe.it.uc3m.es/pub/Programs/, and the project web page is at
>  http://www.it.uc3m.es/ptb/enbd/. Go to www.freshmeat.net and look up
>  "ENBD" for more recent public release announcements.
> 
> It allows you to export a partition or file to another system. On the other
> system you can use it like a local disk. An example is that you could setup
> an enbd as the second device for a mirror and have a backup on another box.

I saw a presentation at OLS regarding this, but the problem is that
it is a block I/O layer, not a network FS layer. So if you want access
to the same data from windows and Linux boxes at the same time, you
need to have your ENBD client export the devices as smb and you've
now added a bunch of latency to all file operations since you now
have:

smb on network -> ext3 on server -> nbd on network -> raw disk on server2

My conclusion after watching the presentation was "why?" I also 
think the same whenever someone says iSCSI, but somebody seems to 
think its a good idea.

~Deepak

-- 
Deepak Saxena - dsaxena@plexity.net 

-> Insert interesting quote here <-