I would like to address the no Single-Point-of-Failure claims that TMS makes in this article.
Mirroring is simply not the same as an N+1 design. It puts the management and recovery outside the storage array which is simply suboptimal. It requires maintaining two solutions - the storage and the external management. Redundancy becomes extremely expensive: two appliances instead of one, double the sparing inside each storage appliance. After replacement, one has to rebuild the ENTIRE storage, where in most cases the failure could be local and requires minimal rebuild.