Article Preview
TopIntroduction
Channel assignment for multi-transceiver wireless mesh networks (WMNs) attempts to increase the network performance by decreasing the interference of simultaneous transmissions. Multi-transceiver mesh routers allow the communication over several wireless network interfaces at the same time. However, this can result in high interference of the wireless interfaces leading to a low network performance. With channel assignment, the reduction of interference is achieved by exploiting the availability of fully or partially non-overlapping channels. Channel assignment can be applied to all wireless networks based on technologies that provide non-overlapping channels. Currently, wide-spread technologies are IEEE 802.11a/b/g, IEEE 802.11n, and IEEE 802.16 (WiMAX). With the relatively low cost for IEEE 802.11 hardware, the number of deployments based on this technology is increasing and channel assignment algorithms are gaining in importance.
Although channel assignment is still a young research area, many different approaches have already been developed (Si, 2009). These approaches can be distinguished into centralized and distributed algorithms. Centralized algorithms rely on a central entity, usually called channel assignment server (CAS), which calculates the network-wide channel assignment and sends the result to the mesh routers. In distributed approaches, each mesh router calculates its channel assignment based on local information. Distributed approaches can react faster to topology changes due to node failures or mobility and usually introduce less protocol overhead since communication with the CAS is not necessary. As a result, distributed approaches are more suitable once the network is operational and running. Another classification considers the frequency of channel switches on a network node. In fast channel switching approaches, channel switches may occur frequently, in the extreme for every subsequent packet a different channel is chosen. The limiting factor for dynamic algorithms is the long channel switching time with commodity IEEE 802.11 hardware, which is in the order of milliseconds. Slow channel switching approaches in contrast, switch the interfaces to a particular channel for a longer period, usually in the order of minutes or hours. Hybrid approaches combine both methods.
An important input for channel assignment algorithms are the particular channel characteristics of the used network technology. For instance, IEEE 802.11b/g offers in theory three non-overlapping channels, e.g. {1, 6, 11}, and all available channels in IEEE 802.11a use non-overlapping frequency spectrums. This means, that concurrent transmissions on these channels should not interfere with each other. In practice, experiments and measurement on different experimental platforms have shown, that the non-interfering characteristics do not hold for many reasons (Fuxjager, 2007; Subramaniam, 2008; Draves, 2004), an important one is the insufficient distance of less than 1 m between the antennas on a single network node. Since mesh routers are usually quiet compact, it is almost impossible to design a multi-radio mesh router with sufficient antenna distance. This is also the case with DES-Nodes of the DES-Testbed, on which the three WiFi antennas are mounted with a distance of about 30 cm. Therefore, we also expect to experience side-effects on the theoretical non-interfering channels, which are subject to experimentation in this article.