A SUMMARY OF MARKOVIAN SOFTWARE AVAILABILITY MODELING

摘要

This paper summarizes stochastic models for software availability measurement. The following three models are reviewed: (1) the basic model, (2) the model with two types of software failures, and (3) the model with two types of restoration actions. In the basic model, the stochastic characteristics of operable and inoperable times are described in relation to the cumulative number of faults corrected perfectly. This model also considers the imperfect debugging environment where the faults causing software failures are not always removed perfectly from the software system. The basic model is modified into the models reflecting customers' operational environments. The second model considers the software failures due to the operational uses deviating from the specification as well. The third model takes notice of the restoration actions without debugging activities. The above models apply Markov processes for describing the time-dependent behaviors of the systems alternating between up and down states. Several quantitative software availability measures are derived from the models.