On Network Model Division Method Based on Link-to-Link Traffic Intensity for Accelarating Parallel Distributed Simulation

Abstract

In recent years, requirements for performance evaluation techniques of a large-scale network have been increasing. However, the conventional network performance evaluation techniques, such as mathematical modeling and simulation, are suitable for comparatively small-scale networks. Research on parallel simulation has been actively done in recent years, which might be a possible solution for simulating a large-scale network. However, since most existing network simulators are event-driven, parallelization of a network simulator is not easy task. In this paper, a novel network model division method based on link-to-link traffic intensity for accelerating parallel simulator of a large-scale network is proposed. The key ideas of our network model division method are as follows: (1) perform steady state analysis for the network model that is to be simulated, and estimate all traffic intensities along links in steady state, (2) repeatedly apply the minimum cut algorithm from graph theory based on the estimated traffic intensities, so that the simulation model is divided at the link that has little traffic intensities in steady state.