A comparison between the tomography based methods and the HMM-Kalman model for the estimation of the traffic matrix of an IP network

This paper deals with a hybrid approach of the Kalman Filtering (Kalman) and the Hidden Markov Modeling (HMM). We have developed a hybrid algorithm between both modeling techniques for accurate estimation of the Traffic Matrix (TM) of a large IP Network (Abilene). This algorithm is entirely devoted to the precise estimation of the Traffic Matrix of the IP Network. The implementation of this HMM-Kalman model had yielded good results on TM estimation. Spatio temporal correlations and other parameters had been calculated and tested. Thus, the main aim of this paper focuses on the comparison of the efficiency of the HMM-Kalman compared to the well-known tomography based techniques. In the work carried out, it appears that the HMM-Kalman has earned the challenge in terms of accuracy allowing at a time the lowest estimation error.