Redundancy optimization of fault tolerant mass storage memory for massively parallel computing systems