作者: J.B. Dugan , M.R. Lyu
DOI: 10.1109/24.370232
关键词:
摘要: This paper presents a quantitative reliability analysis of system designed to tolerate both hardware and software faults. The achieves integrated fault tolerance by implementing N-version programming (NVP) on redundant hardware. considers unrelated faults, related transient permanent imperfect coverage. overall model is Markov in which the states chain represent long-term evolution system-structure. For each operational configuration, fault-tree captures effects faults task computation. parameterized using experimental data associated with recent implementation an NVP current design paradigm. considering typical failure rates coverage parameters. authors results show that it important consider system, since these estimates vary time. Moreover, function for error detection recovery extremely fault-tolerant software. Several orders magnitude reduction unreliability can be observed if this provided promptly. >