Traffic characterization for integrated services networks

A discrete-time queuing analysis is presented for integration of multiple traffic types in a packet-switched TDM (time-division multiplexing) system. The correlation of each traffic type is represented in terms of the power spectral density. The general expression obtained for the aggregate mean queue size indicates that slowly varying traffic types exert the largest effect on the queuing process. Also, a lengthy-steady traffic is highly predictable. An adaptive flow control and routing scheme, based on signal prediction, which successively adjusts the short-burst traffic arrival rate at each TDM node, is introduced and analyzed. The analytical results indicate a substantial reduction in the correlation effect on the system queuing behavior.<
>