Large-Scale IP Traffic Matrix Estimation Based on Fratar Model and ART

Abstract

This paper proposes a novel method of large-scale IP traffic matrix estimation, called FratarArt method by us, which is based on Fratar model and Algebraic Reconstruction Technique (ART). Firstly, we model OD flows as Fratar model and introduce the constrained relations between traffic matrix and link loads. At the same time, we consider the nonnegative constraints of traffic matrix. By Fratar model, we can get a good prior of network tomography. As a consequence, a good estimation of traffic matrix is attained with ART. Finally, we use the real data in true network Abilene to validate our method. The results show that our method can perform fast the accurate estimation of traffic matrix and track its dynamics. In contrast to TomoGravtiy (Zhang et al., 2003), our method improves remarkably and the estimation of traffic matrix is closer to real data.