With the introduction of JBOD support in Cassandra 1.2, is RAID 0/10 still the preferred setup?
It seems that a RAID0 setup may still offer improved read/write performance over JBOD but with the risk of losing the entire node if a disk fails. This isn't a huge issue with R>2 and a large number of nodes.
If a disk fails using JBOD, Cassandra flags the disk, distributes the data to other working nodes, and when the disk is replaced Cassandra would assign new data? (thinking hot swappable drives)