| Dependability analysis is a key step in the design, analysis, and tuning of computer-based systems. Since many safety-critical systems occurring in diverse applications fall under the category of hierarchical systems, the dependability modeling and analyzing of hierarchical systems are thus of great importance. The current approaches to dependability study, such as Markov based and combinatorial approaches, often have various restrictions such as small system size, exponential failure behavior, no repair, perfect coverage, and binary outcome; thus, the distance between the model and the reality of the system is too large. This work is concerned with developing efficient methodologies for modeling and analyzing a wider range of more practical hierarchical computer-based systems with less restrictive assumptions and thus reducing the distance to reality.; In this dissertation, we develop a generalized coverage model to analyze the behavior of the system in the presence of modular imperfect coverage, and provide a general fault tree solution for incorporating modular imperfect coverage into the dependability analysis. We develop efficient combinatorial methodologies for analyzing reliability, performance and component sensitivity of general hierarchical systems, especially, the generalized phased-mission systems with modular imperfect coverage.; A study of maintainable systems is also conducted. We develop computationally efficient approaches that unite the advantages of both combinatorial methods and Markov methods for analyzing complex maintainable system availability and reliability, as compared with the most researchers who construct overall Markov model to analyze repairable systems. Three kinds of maintenance plans are discussed.; The implementation of methodologies for imperfect coverage phased-mission analysis and component sensitivity evaluation into Galileo, a fault tree reliability analysis tool, is also discussed.; Finally, after a summary of major contributions of this dissertation, possible research problems on hierarchical interconnection networks and addressing imperfect coverage using multiple valued decision diagram are suggested as future work. |
