摘要

This study presents an adaptive bandwidth allocation model for multiple classes in wireless broadband network, worldwide interoperability for microwave access (WiMAX) (IEEE 802.16). Data to and from Internet are all conveyed through WiMAX links to its final destination. In order to promise the quality of real-time traffic and allow more transmission opportunity for other traffic types, the proposed adaptive bandwidth allocation (ABA) algorithm would first reserve the unsolicited bandwidth for constant-bit-rate traffic (unsolicited grant services). Then, polling bandwidth is allocated for real-time traffic (rtPS) to meet their end-to-end delay constraints and for non-real-time traffic (nrtPS) to meet their minimum throughput requirements. One of the novelties presented by this study is right in that ABA does not greedily overtake too much bandwidth from the lowest-priority class, best effort (BE). Instead, it is very intelligent to only meet the delay constraint of rtPS and the minimum throughput requirement of nrtPS, while it endeavours to avoid any possible starvation of BE traffic. For the purpose of performance evaluation, a four-dimensional Markov chains is built to analyse the proposed ABA. The analytical results are validated by a simulation. Finally, from the comparison with a previous work, the authors observe the performance superiority of the ABA in satisfying the delay constraints (for rtPS), meeting the minimum throughput requirements (for nrtPS) and reducing the average packet drop ratio (for BE), when traffic environments are varied.