[CentOS] Using (was: Announcing) Gluster Storage Platform

Wed Dec 16 16:13:01 UTC 2009
Les Mikesell <lesmikesell at gmail.com>

Chan Chung Hang Christopher wrote:
> Alan McKay wrote:
>> On Tue, Dec 15, 2009 at 7:48 PM, Christopher Chan
>> <christopher.chan at bradbury.edu.hk> wrote:
>>> A cluster filesystem
>> OK, but you've just given me a circular definition.
>>
>>> When you do not need/want a cluster file system
>> and again ...
>>
> 
> Okay, a cluster/distributed file system that does not have its own on 
> disk format. It makes use of whatever existing filesystem there is for 
> actual storage and allows you to replicate files/load balance requests 
> to files to 'storage servers' of any supported platform.
> 
> At the same time, user level processes on 'clients' access the system as 
> if it was an actual file system.
> 
> This enables one to have Linux clients that run say samba to export the 
> files to Windows clients but the actual files are kept on OpenSolaris 
> servers on zfs. Should the Linux clients all go down, the Windows 
> clients could still access the files on the OpenSolaris servers via samba.

I'm having trouble finding any real information about how (and how well) 
this works and I'd like to know if it would be suitable for a backuppc 
storage archive which generates millions of hardlinks.  Does it deal 
with hardlinks spanning backend storage servers transparently?  And can 
it replicate efficiently enough to have remote copies?

-- 
   Les Mikesell
    lesmikesell at gmail.com