Consider a set of terminal nodes K that belong to a network whose nodes areconnected by links that fail independently with known probabilities. Weintroduce a method for estimating any performability measure that depends onthe hop distance between terminal nodes. It generalises previously introducedMonte Carlo methods for estimation of the K-reliability of networks withvariance reduction compared to crude Monte Carlo. They are based on using setsof edges named d-pathsets and d-cutsets for reducing the variance of theestimator. These sets of edges, considered as a priori known in previousliterature, heaviliy affect the attained performance; we hereby introduce andcompare a family of heuristics for their selection. Numerical examples arepresented, showing the significant efficiency improvements that can be obtainedby chaining the edge set selection heuristics to the proposed Monte Carlosampling plan.
展开▼