The case for architecture-specific common cause failure rates and how they affect system performance

In redundant systems composed of identical units, common cause failures can significantly affect system performance. We show that the probabilities of common cause failure are architecture specific. We derive the probabilities for generalized dual and triplex systems. We analyze and compare three common architectures, i.e., 1oo2, 1oo2D and 2oo3, to determine the effects of common cause failure on system performance as measured by the indices MTTF, MTTFSafe, MTTFDangerous, Probability of Failure on Demand, and availability. We conclude that the 2oo3 provides marginally better availability than the 1oo2D. In all other performance indices, the 1oo2D provides superior performance.