Traffic Matrix Estimation Based on Incomplete Network Link Loads Measurement
Qian Chen, Changda Wang · 2021 IEEE 21st International Conference on Software Quality, Reliability and Security Companion (QRS-C) · 2021
Traffic matrix is an important input requirement to better system management. However, it is resource consuming to improve the estimation accuracy of traffic matrix especially in large-scale wireless sensor networks. The paper proposes a novel method named TMEIM (Traffic Matrix Estimation with Incomplete Measurement), which uses a subset of network link loads to estimate the whole traffic matrixes and therefore decreases the resource consumption. In the TMEIM method, the subset of link loads is first selected by the Bayesian A-optimal design; second, the unknown link loads are complemented through k-element polynomial ridge regression; finally the traffic matrix is estimated through the orthogonal matching pursuit method. In the paper, publicly available Abilene network data are used to test accuracy of the TMEIM method. The experimental data show that with the increasing number of link loads being measured in a network, the accuracy of the proposed method is increasing accordingly.