Reliability prediction of distributed systems using Monte Carlo method

Abstract

Distributed systems consist of layers of switching elements connected together in a predefined topology providing the connectivity between the set of processors and the set of memory modules. Reliability prediction is important as it determines the usability and efficiency of the network to provide services. This paper presents reliability analysis of Shuffle-Exchange Network (SEN) systems using Monte Carlo method with stratified sampling. A SEN, a specific type of distributed systems that consists of 2 × 2 switching elements, has been widely considered a practical interconnection system. Confidence interval of the point estimate is then derived using non-parametric bootstrapping.