Medium access control and routing protocols for wireless mesh networks

Abstract

Wireless mesh networks (WMNs), a.k.a. community wireless networks, have emerged to be a new cost-effective and performance-adaptive network paradigm for the nextgeneration wireless Internet. Targeting primarily for solving the well-known last mile problem for broadband access [1, 2], WMNs aim to offer high-speed coverage at a significantly lower deployment and maintenance cost. As shown in Fig. 4.1, most of the nodes are stationary in WMNs. Only a fraction of nodes have direct access, and will serve as gateways, to the Internet. In addition, several nodes serve as relays forwarding traffic from other nodes (as well as their own traffic) and maintain network-wide Internet connectivity, while the remaining nodes send frames along dynamically selected ad-hoc paths to the gateway nodes with Internet access. WMNs are preferable to existing cable/DSL based networks or wireless LANs (that provide WiFi access), due to the following advantages: (i) mesh networks are more cost-effective as service providers do not have to install a wired connection to each subscriber (20-50K per square mile to establish access, approximately 1/4 of the cost incurred in high speed cable access); (ii) mesh networks are inherently more reliable since each node has redundant paths to reach the Internet; (iii) the throughput attained by a mesh network user can be, in principle, increased through routing via multiple, bandwidth-abundant paths (in contrast, in WLANs the shared bandwidth decreases as the number of users within a HotSpot increases); and (iv) WMNs can readily extend their coverage by installing additional ad-hoc hops. Several cities are planning or have partially deployed WMNs, such as Bay Area Wireless Users Group (BAWUG) [3], Champaign-Urbana Community Wireless Network (CUWiN) [4], SFLan [5], SeattleWireless [6], Southampton OpenWireless Network (SOWN) [7], and Wireless Leiden (in Netherlands) [8]. The academic/research efforts are, on the other hand, represented by the MIT Roofnet project [9], the Rice University Technology for All project [10], and the MSR Self-organizing neighborhood wireless mesh networks project [1]. Although initial successes have been reported in these efforts, a number of performance related problems have also been identified. Excessive packet losses [11-13], unpredictable channel behaviors [11,12], inability to find stable and high-throughput paths [11, 12], and throughput degradation due to intra-flow and inter-flow interference [13-15] are among those most cited. All these problems are rooted in the fact that the notion of a link is no longer well-defined in wireless environments. In network theory and practice, a link is usually characterized by its bandwidth, latency, packet loss ratio and patterns. However, in a WMN, a wireless medium is shared among nodes, and the sharing range is determined by (i) several PHY/MAC attributes such as the transmit power, the carrier sense threshold, and the channel on which an interface sends/receives frames, (ii) intra- and inter-flow interference (which in turn is contingent upon how nodes and traffic are distributed in the spatial and temporal domains), and (iii) environmental factors, such as multi-path fading and shadowing effects, temperature and humidity variation, and existence of objects in between. As a result, all the definitive metrics that characterize a link are no longer well-defined for a wireless link. All the protocols that were devised, and well-suited, for wireline networks will likely yield poor performance or even fail in WMNs. For example, as shown in [11] and [16], the shortest path algorithm with the hop count as the link metric will likely identify paths that are composed of long, lossy links with low bandwidth. To solve (or at least mitigate) the above problems, one should (a) characterize how, and to what extent, wireless links are affected by PHY/MAC attributes and other environmental factors, (b) identify control knobs in the PHY/MAC layers with which the sharing range of a wireless link can be better controlled, and (c) understand the implication of making available these PHY/MAC attributes to higher-layer protocols on system performance optimization. Central to issues (a) and (b) is medium access control (MAC), while issue (c) is usually termed as cross layer design and optimization. In this chapter, we discuss the state of the art in designing and implementing MAC for WMNs in Section 4.2. In particular, we categorize existingMAC-related research into four categories: (1) controlling the sharing range of the wireless medium and increasing spatial reuse; (2) exploiting temporal/spatial diversity; (3) exploiting availability of multiple channels; and (4) exercising rate control. We also introduce in Section 4.3 a modular programming environment, termed as the Transport Device Driver (TDD), that exports the PHY/MAC attributes via well-defined APIs and facilitates cross layer design and optimization, as a case study of cross layer design and optimization. We then present various routing protocols that take advantage of PHY/MAC attributes (such as channels) for route optimization in Section 4.4. We discuss open research issues in Section 4.5. Finally, our conclusion follows. © 2007 Springer Science+Business Media, LLC. All rights reserved.