Exploiting wireless broadcasting nature for high-throughput 802.11 mesh networks

Wireless mesh networking (WMN) has seen great research and commercial interests recently. It is considered a promising technology for implementing wireless community networks. Although the ubiquitous and low-cost Wi-Fi devices make IEEE 802.11 a prevailing choice for wireless mesh networks, the current IEEE 802.11 protocols cannot achieve full utilization of network capacity in wireless mesh networks. When dealing with multi-hop transmissions, IEEE 802.11 MAC presents low efficiency in coordinating concurrent transmissions, ineffectiveness in avoiding interference. In this dissertation, we demonstrate the factors that lead to low capacity in 802.11 mesh netowrks, and we survey the existing work on improving the network throughput. A large class of previous work focuses on coping with the interference, i.e., curbing the negative effect of broadcast nature of wireless signaling. In contrast, another set of work introduced here tries to exploit the wireless broadcasting, and it exhibits promising potential for large throughput gain in mesh networks. We propose three methods that further take advantage of wireless broadcasting. They all aim at achieving high throughput in WMNs, while exploiting the broadcasting nature in different aspects. The first work passively makes use of wireless broadcasting in that it collects wireless link information through mere listening. Specifically, we present a non-intrusive method to model and estimate 802.11 link bandwidth based on radio signal-to-noise ratio (SNR). In a more active way, the second method enables mesh senders to probes their receivers for their MAC statuses so that their following transmissions can be more efficient. Owing to wireless broadcasting, multiple receivers are probed simultaneously with one single probing and their diversity/correlations are speculated. With the diversity information, we propose a smart scheduling strategy. We show that such diversity information can greatly improve throughput of mesh senders. Furthermore, the third method proactively uses the packet redundancy caused by wireless transmissions in local area of a network. Such redundancy are usually ignored or even avoided on purpose by traditional protocols. In contrast, we show that it provides an abundant repository of packets for performing network coding. We propose a new protocol called BEND, which enables each potential forwarder to proactively mix/encode the packets that either are intended to or are overheard by this node. This proactive mixing significantly increases the coding opportunities in the network, leading to high capacity gain.Ph.D.Includes bibliographical references (p. 93-99)by Jian Zhan