A Novel Approach to Large-Scale IP Traffic Matrix Estimation Based on RBF Neural Network

Abstract

Traffic matrix is a key parameter of traffic engineering, which describes the characteristics of IP networks from the global aspects. Though traffic matrix estimation is extensively studied, traffic matrix is generally unavailable in the large-scale IP network and is difficult to be estimated accurately. This paper proposes a novel method of large-scale IP traffic matrix estimation, termed the radial basis function (RBF) neural network and iterative proportional fitting procedure (RBFIPFP) method. Firstly, we model the large-scale IP traffic matrix estimation using the RBF neural network that has been studied widely. By training the RBF neural network, we can build the model of large-scale IP traffic matrix estimation. Secondly, combined with this model and iterative proportional fitting procedure, the good estimations of the large-scale IP traffic matrix are attained. Finally, we use the real data from the Abilene network to validate RBFIPFP. The results show that RBFIPFP can perform the accurate estimation of large-scale IP traffic matrix, and track well its dynamics.