Heavy traffic multiplexing behavior of highly-bursty heterogeneous sources and their admission control in high-speed networks

Abstract

Highly bursty heterogeneous sources, their heavy traffic multiplexing behavior, and their admission control in high-speed networks are discussed. The mathematical model used consists of a discrete-time queue which is fed by the superposition of the nonidentical three-state ternary Markov sources, where each source is characterized by its utilization and the first two moments of its burst period. A simple characterization of the tail of the queue length distribution in the heavy traffic is obtained. This characterization provides a simple metric for call admission control in high-speed networks where the network resources are shared by different sources with highly different characteristics. The proposed metric for call admission control is ideal for real-time operation in a high-speed environment. >