FUZZY RELIABILITY MODELING OF GRACEFULLY DEGRADABLE COMPUTING SYSTEMS

The meaning of failure/success is hard to determine for gracefully degradable computing systems because of the attribute of graceful degradation. In this paper, the concept of fuzzy success/failure (state) is introduced to represent the system structure, the performance and other considerations (e.g. cost). Thus, the transition from fuzzy success to fuzzy failure is viewed as a fuzzy event. Accordingly, the probability that no transition from fuzzy success to fuzzy failure occurs in a predetermined time interval is defined as fuzzy reliability. Fuzzy reliability can be physically interpreted as the probability that no substantial performance deterioration occurs in a predetermined time interval. Also proposed for systems with fuzzy states are measures including system fuzzy availability, fuzzy unavailability, stationary fuzzy reliability, virtual random lifetime and mean time to fuzzy failure. Numeric results show that all the fuzziness, coverage factor, and maintenance have significant effects on the fuzzy reliability behaviour of gracefully degradable computing systems.