Probabilistic analysis of real-time dependable systems

The design of real-time dependable systems is dominated by the need to meet real-time deadlines. The authors present a methodology for analyzing durations an real-time dependable systems and for estimating the probabilities that real-time deadlines will be met. Probability density functions for the durations of composite operations are obtained by convolving the probability density functions for the durations of simpler operations. These probability density functions may be correlated, and such correlations must be handled correctly when the convolutions are performed. Example applications of the methodology are given for the Totem multicast group communication protocols.