*Distributed* Filesystem (was Re: [clug] Clustering Filesystem)

[Alex Satrapa]

On 23 Aug 2005, at 22:39, Paul Wayper wrote:

> At 09:45 PM 23/08/2005, [Michael Cohen] wrote:
>
>> Hi List,
>>   I have recently installed a bunch of machines on my network.  
>> Each of
>>   these has a sizable hard disk on it. I am after a clustering  
>> network
>>   filesystem that has the following properties:
>>
>
> I've done a little bit of research into this, but I haven't  
> actually played with any of them.  The best developed, as far as I  
> can see, is Coda

"Clustering" file system would imply something like GFS (the "Global  
File System") which is a file system used by processor clusters,  
typically when talking to a SAN.

To spread your file system out between machines that are located away  
from each other you want a *distributed* file system.

Note that Coda relies on a "master" server, with clients replicating  
changes back and forth. I'm not sure that it's what you want in order  
to make use of that extra hard drive space, though it might be  
possible to wrangle one machine to be a client of the others and run  
some kinda of JBOD in software over the various Coda file systems.

My (admittedly poorly developed) understanding is that a Coda client  
will eventually end up with a copy of the entire file system - thus  
Coda is a distributed/replicating file system. Samba + smbfs is not  
an option, in my experience - smbfs squashes all the ownership and  
permissions to be rwx by the user mounting the volume.

I'll be playing with Coda a little over the next few weeks (once I  
get these PITA P5GD1 motherboards to work for me), so feel free to  
contact me off-list if you want to compare notes.

Alex Satrapa             M: +61 4 0770 5332
grail at goldweb.com.au     W: http://homepage.mac.com/alexsatrapa