[ale] Fault Tolerant High Read Rate System Configuration

[Ed Cashin]

It's probably too wild for this case, but I wonder why folks aren't
using PVFS2 in situations like this.  I'm guessing it's not quite
mainstream enough.

The promise of PVFS2 is to provide highly scalable I/O in a
way that's pretty simple at the root of it: do RAID over the network
between a bunch of computers, each with its own locally attached
storage.  In PVFS there wasn't as much redundancy, so it was
more limited to checkpointed academic research, etc., but PVFS2
is integrated with a lot of well-tested safeguard technology like heartbeat.

-- 
  Ed Cashin <ecashin at noserose.net>