Dependability evaluation of the saturation technique

An evaluation is presented of the dependability of a technique, called saturation, for improving fault tolerance in distributed systems. The evaluation is based on simulation and on Markov processes. First, comparative performance results are obtained by simulation of the saturation algorithm described in Estelle, and second, a Markov model is established in order to calculate the failure rate of a redundant task set using saturation. Some of the parameters of the Markov model are derived from the simulation. The saturation technique is compared to classical majority voting techniques.<<ETX>>