Dependability analysis of large-scale distributed systems using stochastic Petri nets

Dependability models of complex distributed systems using Markovian techniques suffer from state space explosion. Several methods for controlling the state space explosion problem have been proposed in the literature. These largeness avoidance methods include state truncation, model composition, behavioral decomposition, time-scale decomposition and fixed-point iteration. In this paper we briefly review these different methods in the context of dependability evaluation of large-scale distributed systems. We use the stochastic reward nets (SRN) as the modelling formalism, and show how the different methods can be implemented using the various structural constructs available within SRN models. We illustrate the application of these methods by an example.