Static Channel Assignment in Multi-radio Multi-channel 802.11 Wireless Mesh Networks: Issues, Metrics and Algorithms
A. K. Das, S. Roy and R. Vijaykumar
Book chapter, “Wireless Mesh Networks: Architectures, Protocols and Standards”, CRC Press, 2006 [to appear].
Download: PDF
Introduction: The emergence of cost-effective wireless access technologies to mobile end-users such as 802.11 has changed communications and computing in significant ways. Its success to date has been largely due to its deployment in the home and small enterprise segments and various ‘hot spot’ scenarios where it has limited coverage and serves only a few users simultaneously. Currently, there is considerable interest in expanding 802.11 networks to large-scale enterprise scenarios to provide wide-area wideband access to a significant number of users. This requires a proliferation of access points (APs) over the desired coverage area. Wide-area coverage using 802.11 BSSs should naturally look to principles of cellular systems engineering for successful scaling. The key to one-hop capacity scaling (such as between a client and AP) is based on frequency reuse spatially. Given any number of orthogonal channels, neighboring APs are assigned the available orthogonal set in a systematic manner (e.g. the familiar frequency reuse patterns in cellular networks). The resultant signal-to-interference + noise ratio (SINR) at the edge of the ‘cell’ (or BSS) along with the inherent properties of the 802.11 Distributed Coordination Function (DCF) protocol then essentially determines the throughput per cell obtainable. Scaling the aggregate network throughput (over the coverage area) is directly related to reducing the re-use distance between co-channel APs without degrading the SINR at the cell edge (equivalently, increasing the spatial reuse factor). This can be achieved by a combination of approaches (notably, among others, the use of Directional Antennas for beamforming at the APs, which we will not consider) – the most obvious being the availability of increased system bandwidth (equivalently, more orthogonal channels). Currently, only a very limited number of such orthogonal channels are available: 3 in 802.11b (2.4GHz) and 12 in 802.11a (5 GHz). Although greater worldwide allocation is anticipated for unlicensed use in the future, it is clear that relying primarily on increased available bandwidth for scaling is not a feasible option. Accordingly, for any given system bandwidth, optimizing network performance necessarily requires improving the entire protocol stack. A promising option for scaling the capacity of a wireless access network is to configure a Layer-2 mesh as is being currently planned within the IEEE 802.11s Task Group. This implies a direct wirelessly inter-connected set of mesh nodes (which comprise Access Points that allow direct client access and future ‘routers’ that only relay packets between other mesh elements) to form a multi-hop network. The ad-hoc (but static) nature of mesh node deployments results in a significant spatial variability of the multi-access interference (MAI) seen at any node location, implying variable location-dependent node throughput. Hence, effective topology modification mechanisms including power control, node clustering and channel assignments are anticipated to be important design degrees of freedom.