Network performance reliability evaluation based on network reduction

Abstract

Network reliability is very important in our modern society. The primary network reliability research focuses on connectivity reliability. Recently, network performance is receiving more and more attention, and network performance reliability concept is promoted, such as travel time reliability. But it is very hard to evaluate it. One of the reasons is high computational complexity. Although this problem also exists in connectivity reliability, it is much worse in performance reliability, because it concerns capacity and traffic in multiple states. In order to reduce complexity of evaluation, a network performance-reliability-preserving reduction method is proposed in this paper. Firstly, we define the network performance reliability concept. It is a probability that performance indicators remain their values within expected ranges under a certain traffic flow. Performance indicators include traditional performance parameters, such as loss rate, throughput, and delay. Secondly, we propose a network performance reliability evaluation method which combines Monte Carlo reliability simulation with a network performance model. Thirdly, we propose a performance-reliability-preserving reduction method on the basis of network graph transformation methods, which can reduce the complexity of our performance reliability evaluation method by scaling down networks. A case study shows a very good correspondence of performance reliability between the simplified and initial networks.