OpenAFS, DRBD

[tmetro-blu-5a1Jt6qxUNc at public.gmane.org (Tom Metro)]

Richard Pieri wrote:
> OpenAFS on, well, any Unix of your choice beats ZFS for all of the
> above.  ZFS's snapshot and clone mechanisms are straight from AFS.
> With redundant OpenAFS file server machines you'll find it hard to
> beat it for reliability.

I tend to think of AFS as being at the network layer, same as NFS, while
ZFS operates at the block layer, with some network-layer features. Does
AFS also go down to the block-layer?


> On the other hand, OpenAFS needs more work to initially set up, and
> it requires a Kerberos realm. 
...
> It's designed for large scale university environments: lots of
> concurrent users with lots of files on lots of nodes distributed
> around a large geographic area.
...
> OpenAFS is overkill for a small shop with a handful of users.

I'd like to see something with AFS-style features (particularly the
security), that was optimized for a small, 2-node setup. (If you rule
out SMB and NFSv3 for security reasons, the challenge becomes finding
client side support.)

In my other post I mentioned Ceph, a new distributed file system, but it
also seems optimized for many nodes, not 2.

A typical small business could benefit from a simple setup with a pair
of redundant filers that can load balance and automatically fail over.
If implemented right, you'd also forgo using RAID on the individual
filers (other than perhaps striping for performance), reducing the disk
cost, and shifting that money into overall system redundancy.


Anyone tried using DRBD (http://www.drbd.org/) to build this sort of thing?

 -Tom

-- 
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/