[lug] Gluster question -- reframed

[Vince Dean]

At a BLUG meeting a few months ago, I asked the group for opinions about
the Gluster distributed file system.

I wanted to provide:
- replication to guard against hardware failure at a node
- convenient fail-over if a server fails or goes off-line

Of less immediate importance would be:
- the ability to combine disk resources from several servers and
  make them look like a single larger file system

The context was a somewhat chaotic environment including both stable,
professionally managed servers and desktop machines which might be
turned off or rebooted at any time.

In that context, the response was negative:  gluster would not be a good
idea.

Suppose I change the context so that all the machines are nice stable
servers which run in machine rooms and are not subject to arbitrary
interference from users.  In that context, is gluster or any other
similar distributed file system well-enough established that I should
trust it with research data which would be quite difficult to reproduce?

The servers might each have a single RAID with 10 to 30 terabytes of
space.

Thanks,
Vince