Rate Allocation And Admission Control For Differentiated Services In Cdma Data Networks

In this paper, we propose a unified framework for rate allocation and admission control for CDMA data networks with a goal of maximizing the service provider\u27s revenue taking into consideration user churn behavior, i.e., users\u27 migration from one provider to another. The power constrained discrete rate allocation (PC-DRA) problem that finds the rates at which the base station must transmit to each user is very difficult because of the nonlinear nature of the objective function, the discrete nature of available rates, and the complex relationship between data rate, the target SINR, and the total power constraint. We prove that the PC-DRA problem is NPComplete for CDMA data networks and then propose three heuristics - a two step selective rate reduction (SRR) scheme, a genetic algorithm (GA), and an improved genetic algorithm (IGA) to solve it. We show that all three heuristics can achieve suboptimal that is very close to the global optimal, while SRR provides solutions of high quality with much less computation time than GA and IGA. At the frame level, the rate allocation maximizes the service provider\u27s revenue by adjusting individual user\u27s data rate on a frame by frame basis. On the other hand, the service provider\u27s revenue varies with the system load, which is decided by admission control at the session level. Thus, the admission control needs to maintain an optimal system load. An efficient resource management framework is provided by the cross level (layer) interaction between the link layer (rate control) and the network layer (admission control) in the sense that 1) they share the same objective of maximizing the service provider\u27s revenue, and 2) the solution to the rate allocation problem provides critical information for the admission control decision. The proposed framework is evaluated by extensive simulation experiments with realistic Web browsing and FTP traffic models. Simulation results confirm that the framework is able to provide class-based differentiated data services that also maximize the service provider\u27s revenue. © 2007 IEEE