Design of Multi-tier Wireless Mesh Networks
Raghuraman Rangarajan · 2009
In this thesis, we investigate the issue of automated design of capacity-constrained Wireless Mesh Networks (WMN). We argue for the necessity of applying network design methodologies from wired and cellular network fields in wireless network design scenarios and present algorithms for Wireless Local Area Networks (WLANs) and backbone topology design. The deployment scenario we envision is a campus of office buildings requiring wireless connectivity. The client nodes to be deployed in each office, their application traffic requirements and the deployment layout are given. We identify three main stages in the design of such wireless networks: 1) Association of clients to access points, 2) WLAN topology construction and 3) Backbone topology construction. Capacity provisioning and network cost minimisation are the two constraints imposed on the design problem. In the first stage, we define the AP-assignment problem, that is, the problem of associating client nodes with the nearest Access Point (AP) and investigate various access point association scenarios. We compute the performance of 802.11 WLANs under homogeneous realtime application deployments with theoretical and OPNET simulation results for various voice and video codecs. We capture the performance, in terms of number of flows supported for an application, as the capacity of a WLAN. We then examine heterogeneous application deployments and show the inability of 802.11 DCF mechanism to handle them. We propose an extended DCF for handling scenarios where delay-sensitive and delay-tolerant applications are deployed together. Next, we propose a novel approach called sub-optimal APassignment and show that the utilisation of 802.11 DCF can be increased by up to 75%. We show how the solutions to the AP-assignment problem can then be used for abstract representations of the 802.11 DCF MAC in wireless design problems. In the second stage, we define a network design problem for constructing WLAN topologies. The scenario we consider is intra-office connectivity for client nodes. We present a recursive bottom-up algorithm for capacity-constrained topology construction. The topology construction algorithm considers the deployment scenario, the client nodes deployed and their application scenarios as inputs. The APassignment solutions are used as a construction mechanism in the form of link specification functions. We introduce a new object, called composite unit, as an abstract building block for network topology construction. The generated topology is then validated with simulations (OPNET Modeler). In the third stage, we define a wireless mesh network (WMN) design problem for constructing a mesh topology for a campus-like scenario. The design problem is defined as a case of traditional network design problem for optimal node location and topology construction. These problem form Mixed Integer Linear Programming (MILP) formulations and are solved with an MILP solver (CPLEX). We have built a tool to implement our multi-tier wireless design solution. The tool, the Wireless Infrastructure Deployment tool (WIND), designs topologies for both WLANs as well as WMNs. WIND takes information about nodes deployed, their properties and deployment layouts to construct logical network topologies. The input parameters and output topology use XML schemas compatible with data formats of the OPNET Modeler network simulator. This allows the topology to be input directly to the simulator for validation. Using simulation we show that the constructed topologies satisfy the given constraints on application scenarios, protocols and deployment scenario. We present case studies of constructing topologies for both WLAN and WMN scenarios.