| Wireless mesh networks (WMNs) have gained considerable interest in expanding IEEE 802.11 networks to large-scale enterprise and community scenarios, primarily because its capability of providing wide-band ubiquitous network access to a significant number of users. However, the mesh nodes are typically limited in resources such as bandwidth, computation power and memory. Moreover, the shared wireless medium introduces the interference. All these result in unsatisfactory network performance. In this dissertation, we tackle several issues related to the system performance of the wireless mesh networks. In the context of the network capacity enhancement, we focus on using the techniques of multi-channel multi-radio (MCMR) and partially overlapped channels (POC) respectively; In the context of network reliability, we focus on the measurement-based analysis and prediction of reliability in an experimental way.;Capacity limitation is one of the fundamental issues in wireless mesh networks. We start our investigation by addressing capacity improvement enabled by multiple radios and channels with a centralized scheduling approach. In a MCMR network, Our objective is to find both dynamic and static channel assignments and corresponding link schedules that maximize the network capacity. We focus on determining the highest gain we can achieve from increasing the number of radios and channels under certain traffic demands.;When the number of orthogonal channel is few or multiple radios are not available, the use of partially overlapped channels is explored to enhance the capacity in wireless mesh networks. We investigate how the interference can be useful for communication when POC is properly utilized. We propose novel channel allocation and link scheduling algorithms in the MAC layer to enhance network performance. Due to different traffic characteristics in multi-hop WMNs compared to those in one-hop 802.11 networks, we perform our optimization based on end-to-end flow requirement, instead of the sum of link capacity. In addition, we discuss other factors affecting the performance of POC, including topology, node density, and distribution.;To better understand how reliable a wireless mesh network performs, we perform a systematic experimental study to investigate the impacts of different factors on the unreliability of a mesh network and the sources causing such unreliability. The factors that we studied include traffic load, number of hops and flows, transmission rates, maximum retransmission limits and the RTS/CTS mechanism. Our results are based on measurements performed on our real-world mesh network testbed. In identifying the sources of packet loss, we developed a tool, FlowPaC, to collect flow-based statistics at different points in the system to understand the effects of the MAC layer parameters and the traffic attributes. We also explore the potential remedies for the system configuration and thereby improve the reliability of wireless mesh networks.;Prediction of network reliability of WMNs at the level of individual flows in a small time granularity becomes very important, but is missing in the literature. Traditionally, the reliability of WMNs is measured as long-term averaged metrics, such as packet delivery ratio. However, WMNs are very dynamic in nature due to fluctuation of channel conditions and contention among multiple flows, so long-term averaged metrics cannot reflect the short-term behaviors of the network. Moreover, the users may require a sustained reliability for a certain period of time. In this dissertation, we measure reliability using metrics of both the goodput and packet loss in short intervals of time for individual flows. We propose a model to predict the short-term reliability for a user's flow in WMNs. Based on the measurement results obtained from our tools Rater and CalMedium, our model can achieve the prediction accuracy and avoid the complexity of the inter-dependency among the traffic, different system queues, interference, and MAC-induced factors. The predicted results have a good match with those obtained from our real-world mesh network testbed for both single-flow and multiple-flow scenarios. We also demonstrate two application examples of utilizing our work: finding the bottleneck rate of a flow and admission control. |
