Distributed Resource Allocation in Contention-Based Wireless Networks

Due to the limited bandwidth in wireless networks, careful resource allocation schemes must be used to allocate bandwidth to applications. There are two typical types of applications in wireless networks that require resource allocation: soft realtime applications and best effort applications. While soft realtime applications require QoS support for effective bandwidth utilization, best effort applications can adapt to changes in bandwidth and are more concerned about network utilization. Due to this difference in application requirements, intelligent resource allocation schemes must be used to provide the desired amount of resources to realtime flows while at the same time allocate the rest of the bandwidth to best effort flows efficiently. The goal of this thesis is to explore two different methods to support such intelligent distributed resource allocation in contention-based wireless networks. The first method uses signaling between nodes to coordinate resource allocation. The benefit of this {\em signaling approach} is that it does not require any specialized MAC layer and is based on existing scheduling algorithms used in current MAC layer protocols such as IEEE 802.11. Due to the shared nature of wireless medium, a signaling approach must effectively coordinate resource allocation among interfering nodes. This thesis addresses two challenging aspects of designing such a signaling approach: efficient signaling and accurate estimation of available bandwidth. The second method that relies on advanced MAC layer scheduling and MAC layer resource monitoring. Such a {\em MAC layer approach} is able to achieve low message overhead and can easily adapt to the dynamics of the network. For the MAC layer approach, this thesis includes four novel MAC layer scheduling algorithms based on dynamic contention window control: DDA for delay assurance to realtime traffic, DBA for throughput assurance to realtime traffic, BEP for rate policing of best effort traffic and GCA for flexible and efficient bandwidth allocation of best effort traffic. This thesis also describes a novel framework, QPART, that provides reactive admission control for a MAC layer approach. All these schemes are fully distributed and can support a variety of resource allocation policies without requiring communications among neighboring nodes