Hi John

> > http://www.kip.uni-heidelberg.de/ti/ClusterRAID/
> > This is a highly stable, distributed, redundant file system using endb...
> Bruce,
>  this looks interesting. I cannot see a good overview on the site
> though. I'm familiar with using drbd  http://www.drbd.org
> At first look this seems to be using a similar endb network block device
> to have RAID 5 storage with the drives being located on different
> machines on the network. Maybe I have picked this up wrong.
You understood perfectly. I am not involved in the clusterRAID
project, except that we collaborated with them previously. I know that
they have deployed it on a cluster succesfully with a 2.4 kernel with
arrays of IDE disks iwth USB interfaces to the mainboard. This meant
that the disks were hot-swappable. (USB was used, because at the time
the SATA disks weren't in widespread use; but this would be a better
option, they had some problems with the USB connectors)

> Also the 'precise accounting patch' is available for 2.6.7
> Is this necessary for using the filesystem?
I think so. As far as I know, the software is stable for the 2.4
kernel, but I'm pretty sure that someone in Heidelberg is working or
has worked on porting it to 2.6. This filesystem was developed for
storing petabytes of data reliably on cheap hardware, hence the
requirement of reduncancy, "distributedness" and hot-swapping.

If you're really interested, you could send a mail to Arne Wiebalck
who did his PhD on it [log in to unmask] and I'm sure he
would be able to give you more information.

Cheers
Bruce

--
Bruce Becker
DAPNIA/SPhN/CEA Saclay
Fr-91191 Gif sur Yvette Cedex
tel :(w) +33 1 69087482 | (m) + 33 62 739 5720 | (f) +33 1 69087584
IM  : AIM/Jabber - brucellino | Yahoo - uctbruce
WEB :http://hep.phy.uct.ac.za/~becker "Viel hilft viel"