Traffic description using spectral characterization of wide-band input processes in ATM networks

Development of multimedia input traffic models has heavily relied on laboratory experiments of individual sources like voice and video. Such models are therefore ad hoc and are generally limited in validity to the range of conditions for which the experiments have been carried out. The authors explore the use of a new technique of spectral characterization of wide-band input process in asynchronous transfer mode (ATM) networks. The technique is based on the eigenstructure of input Markov chains (MC) which helps localize wideband sources in a subspace. Assuming the input process is stationary and properly sampled at each measure point, taking the average over a sufficiently long time interval will result in two input vectors: the power spectrum and the phase spectrum. The authors propose a new traffic descriptor for input sources based on the power spectrum and the phase spectrum and demonstrate that such a descriptor is sufficient for efficient network control and resource management. They also propose a method of frequency-domain analysis of input sources characteristics at each network entry point and intermediate node