Storage systems are built of fallible components but have to provide high degrees of reliability. Besides mirroring and triplicating data, redundant storage of information using erasure-correcting codes is the only possibility to have data survive device failure. We provide here exact formula for the data-loss probability of a disk array composed of several RAID Level 6 stripes. This two-failure tolerant is not only used in practice but can also provide a reference point for the assessment of other data organizations