[ Thanks to Paul Shread for this link. ]
“The Gluster Storage Platform doesn’t use a metadata server to remove a common I/O bottleneck. The company says this approach provides “true parallel data access, ensures linear scalability and eliminates the risk of metadata server failure or corruption.”
“The clustered NAS system also providers a unified global namespace that the company says can scale to hundreds of petabytes in a single volume across multiple storage nodes. Data can be replicated or mirrored for high availability, and the software also performs error detection and correction within files while they are running and during recovery from hardware failures”