An approximation algorithm for optimal resource allocation in multi-service broadband networks

In this paper, we consider a multi-service wireless ATM network carrying different types of traffic. The optimal resource allocation scheme is designed to obtain an efficient and fair allocation among different traffic classes. The Markov decision process analysis is most often used in performance evaluation. Although exact analyses by assuming certain traffic models are available in some cases, the computational efforts required are prohibitive even for a small size network. In consideration of numerical tractability, we derive an approximation recursive algorithm which enables us to deal with large size network systems. The ideas behind the approximation include the separation assumption and multi-dimensional state space mapping into one-dimensional space. This greatly reduces the complexity of analysis while yields reasonable approximated results. Simulation results are presented to illustrate the accuracy of approximation.