Cross correlation analysis and construction of joint distribution traffic model

Traffic models play an important role in network design and performance evaluation. Since the usual assumption of independencies between different traffic streams, these models are not able to precisely reflect the characteristic of real network traffics. In this paper, we present a joint distribution traffic model which can express the dependency of two traffics. This model is constructed with parameters of self correlation and cross correlation. In additional, we can predict the traffic data in a real network according to another relative traffic. We do many simulations in different ports of CERNET, and shows that our model gets well accuracy to the real traffic data.