Chan Chung Hang Christopher wrote:
Alan McKay wrote:
On Tue, Dec 15, 2009 at 7:48 PM, Christopher Chan christopher.chan@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?