Cross-layer solutions for traffic forwarding in mesh networks

Abstract

Wireless mesh networks [1] is an emerging wireless technology that allows robust and reliable wireless broadband service access at relatively low cost. They include two types of nodes: mesh routers and mesh clients. Both types of nodes operate not only as hosts but also as routers, forwarding packets on behalf of other nodes that may not be within direct wireless transmission range of their destinations; in addition, a mesh router may have gateway/bridge functionalities [1]. Wireless mesh nodes dynamically self-organize and self-configure, automatically establishing and maintaining mesh connectivity among themselves. This chapter focuses on a mesh network using the IEEE 802.11 technology. Consider the network section including mesh clients and routers (hereinafter also called nodes) that wish to connect to a mesh gateway, through either direct or multihop communications. The problem addressed here is how to transfer traffic between the wireless nodes and a gateway in a fair, efficient manner. Routing protocols for wireless networks are usually designed considering that all nodes within transmission range of a transmitter are equivalent. However, this is often false, as the quality of the channel toward (and from) different one hop neighbors may significantly vary with distance, presence of obstacles and interfering transmissions. Also, since 802.11 off-the-shelf devices implement rate-adaptation techniques, the link quality directly determines the data transmission rate to be used between pairs of nodes. Thus, measuring routing distances in terms of number of hops may be misleading, as routing through a larger number of high-rate hops may lead to a higher network throughput with respect to performing fewer low-rate forwards [2,3]. An example of this behavior can be observed in the simple scenario of Fig. 10.1. Here, all nodes are within receive range of each other, and both A and B send CBR over UDP data to C using the 802.11 Distributed Coordination Function (DCF) to access the radio channel. Assume that A enjoys an optimal channel quality toward C, while B experiences a low quality channel due to distance and/or obstacles. The resulting decrease in B's data transmission rate is the cause of the well-known 802.11 anomaly phenomenon [4], which reduces the overall network throughput, as shown in Fig. 10.2. However, if the channel quality between B and A is good enough and A relays data from B toward C, then the resulting overall network throughput is better than that obtained with a 2 Mb/s direct data transfer between B and C. Even higher improvements are obtained when employing TCP, which introduces reverse acknowledgments flows from C to A and B. An idea to overcome the problem described above is to design a routing protocol accounting for medium access control (MAC) and physical layer performance in the route computation, by making multiple fast hops preferable to single slow ones. The joint design of MAC and routing schemes is however a challenging issue. Indeed, even if the use of relay nodes may alleviate the anomaly effect and increase the system throughput, as shown in the example above, the role of relay is a thankless one: in addition to its own traffic, the relay node must carry other nodes' traffic. Therefore, some incentives are needed so that the throughput of the relay node is close to that of the same node without relay traffic. In this chapter, we first define and examine two relay strategies that aim at giving relay nodes some incentives for their roles, while at the same time enhancing the overall throughput of the network. The first strategy involves the implementation of two Logical Link Control (LLC) queues at each node: one handles 'local' traffic, the other collects relay traffic from other nodes. The second technique relies on the enhanced distributed channel access (EDCA) specified by the IEEE 802.11e draft standard [5]. We then use experimental measurements and simulation results to derive some guidelines on designing an efficient, cross-layer, relay selection scheme that accounts for quality and transmission rate of the available links. We describe a relay selection algorithm and define a relay-quality aware routing [6], as an extension of the Optimized Link State Routing (OLSR) [7] protocol. The rest of the chapter is organized as follows. Section 10.2 reviews some work on routing in ad hoc networks. The benefits of multihop forwarding in counteracting the anomaly effect are discussed in Section 10.3, with the help of experimental measurements. Section 10.4 describes the proposed traffic forwarding strategies and their performance. Section 10.5 summarizes the major lessons learned from experimental measurements and simulation results, and presents our relay selection algorithm. Section 10.6 describes its implementation and shows some performance results. © 2007 Springer Science+Business Media, LLC. All rights reserved.