Designing for Network and Service Continuity in Wireless Mesh Networks

Abstract

The paradigm of wireless mesh networks has matured in the last decade after extensive research and development. As opposed to the time of their inception, current mesh networks are expected to deliver carrier-grade access services to their users. With the increasing expectations, it is generally recognized that there is a lack of understanding of the level of service continuity that mesh networks can provide to its users. In this research, we address a set of network design problems and propose solutions to improve the service continuity of wireless mesh networks. First, we investigate power control as a network layer problem in mesh networks. It is commonly acknowledged that reducing transmit power levels of nodes to the minimum required to retain connectivity always increases network capacity. In this work, we show that increasing power level of nodes can be beneficial in many cases depending on network’s traffic and topological characteristics. With extensive analysis and simulations, we characterize achievable spatial reuse and capacity with respect to power control in various practical scenarios. Second, we address the problem of generation of traffic hot-spots in wireless networks, and propose a novel power control scheme for load balancing. Using a heuristic based on the concept of centrality, we show that if we increase the power levels of only the nodes which are expected to relay more packets, significant relay load balancing can be achieved even using shortest path routing. Different from divergent routing schemes, such load balancing strategy is applicable to any arbitrary topology and traffic pattern. With this understanding of power control and load balancing, we investigate the service continuity issues of an urban-scale mesh networks. We define and study two essential service continuity metrics - connection robustness (K-center availability) and performance robustness (K-center performability). K-center availability is the fraction of the time a mesh network is available to its users, while performability is a composite measure of performance (here - throughput) and robustness that captures aggregate experience of mesh users over a long period of time. We develop efficient algorithms for estimating these quantities for large urban-scale meshes and study their characteristics in various existing mesh networks. Our study reveals numerous novel insights that can help in designing mesh networks with high availability and performability. While evaluating the performability, we used ideal case throughput performance estimation that ensures fair network resource allocation among the flows. Although this allows faster computation which is especially necessary for large state space exploration, it is difficult to implement in real-world using CSMA/CA MAC. To address this, we target the issue of unfair and lower utilization of network resources in wireless networks. Specifically, in a mesh network, nodes closer to gateways often achieve a larger share of bandwidth resources while nodes farther away often starve for useful bandwidth. Using back-pressure utility maximization framework, we provide a solution to the issue of spatial bias that can be implemented using current 802.11 standards. To further increase the efficiency of back-pressure policy in mesh networks, we propose a dynamic packet aggregation scheme that is especially effective in Internet-like packet size distribution. Along the same lines, we use the back-pressure invariants to develop a variable channel width assignment scheme that can yield maximum benefits of using varying-sized channels while remaining fair in network-wide resource allocation.