Model-Driven Dependability Analysis of Virtualization Systems

After many years of silence, virtualization recently became a hot topic once again, especially in the field of system dependability. Current researches mostly focus on more technical methods without taking into account the problem of how to verify the dependability of these technics. This paper proposes a Markov model to describe virtualization systems and gives the solving methods to quantitatively analyze their dependabilities including reliability, availability and maintainability. Through the numeric results, we get the relationship between the specific dependability evaluating attributes of virtualization systems and the characteristics of virtualization systems' components. Then evolve some conditions which should be satisfied by the virtualization for enhancing system dependability, including the number of backup virtual machines, the failure and repair rate of hypervisor and so on.