Stochastic bounds for microprocessor systems availability

The computing of stochastic bounds has become an efficient technique to obtain performance predictions for computer systems by the mean of Markovian models. However, the quality of these bounds may be affected by several properties related not only to how to use the technique but also to the Markovian model itself. On the other hand, multiprocessor systems have become an efficient and widely used infrastructure to run several critical-life applications. In this paper, we describe how to calculate a stochastic bound for the multiprocessor system availability. We focus on the irreducibility of the model to show how it might influence the quality of bounds.