Re: [tlug] distributed file systems

[Christian Horn <chorn@example.com>]

On Mon, Feb 15, 2010 at 06:01:15PM +0900, Sach Jobb wrote:
> 
> Also it seems that Sun thinks that lustre (http://wiki.lustre.org) is
> the answer to this problem.

Has anyone experiences with lustre on wide areas?
What i recall from your usecase this was for widely distributed servers.
Lustre is optimized for high performance, also having a file accessed by
multiple servers at one time.  

For distributed filesystems there are also GFS, GPFS and OCFS2, but all
optimized to operate on the same blockdevice and having all nodes
interconnected with low latency - so not widely spread locationwise.

A shiny Porsche for widespread distributed filesystem could be AFS,
used by some universities.  Not trivial to setup, relies on Kerberos for
authentication, but has nice features like automatic caching on nodes 
near to places where data is accessed and automatic redundant storing of
data as in "make sure this data is stored twice somewhere".


Christian