A new call admission control scheme based on the self similar nature of multimedia traffic

The self-similar behavior of multimedia traffic was confirmed by several experimental measurements over LANs, MANs and WANs. These statistical properties were also pointed out in VBR coded video sequences and in signaling traffic over N-ISDN networks. The long memory properties of the stochastic processes, used as mathematical models of real traffic, need new architectural approaches in designing the functionalities of a B-ISDN network. We consider fractional Gaussian noise (fGn) as a traffic model completely specified by its mean /spl mu/, variance /spl sigma//sup 2/ and Hurst parameter H, which expresses the degree of self-similarity of the traffic pattern. The paper presents a novel approach for a call admission control scheme which employs the Hurst parameter as an index of burstiness suitable to characterize the selfsimilar nature of multimedia traffic it assumes also a knowledge of the mean and the variance of the process. These parameters can be obtained with an off-line statistical analysis of the traces (in the case of distribution services such as video on demand) or by real-time suitable estimators (in the case of interactive services such as videoconference). The analysis of the multiplexing gain obtained considering a fair distribution of the bandwidth between homogeneous and heterogeneous sources (with respect to their Hurst parameters) is presented. The results are compared to peak allocation schemes currently adopted over wide area ATM networks.