I read this on Slashdot a couple of days ago and found it pretty interesting … RAID is essentially dead because of large disks.  In the time it takes to reconstruct a failed drive, there will be another unrecoverable read error in one of the other disks, essentially making the reconstruction fail, which means that the entire RAID system is dead — you’ve lost your data and cannot recover it.  Intuitively it feels like we’re going to have the same problem with JBOD eventually — we won’t be able to rebuild a disk on a failed server before another drive fails.  The only way to protect against this is to have some system that can rebuild from multiple drives, some of which have failures.  Essentially, we have to assume that there will be parts of every disk that are bad and merely talk to the parts of the disk that are working.