Resource and Traffic Management Techniques for Performance Optimization in Wireless Mesh Networks

Since Internet was born, it has steadily evolved. New devices, new needs, new applications, new protocols, and this evolution is not going to slow down in the near future. As a result of the introduction and standardization of wireless access technology, we are living through an inevitable replacement of the old wired technology. We are going through a wireless revolution, which brings with it new ways to access information: simply, anywhere, anytime. The new Internet challenge is to embody the new generation of wireless technology, giving users mobility and ubiquitous access. Wireless technology allows to extend Internet to rural areas, building blocks, home devices. Mesh networking did grow up and did become a solid technology, and it is ready to be widely adopted. Wireless communication, though, is affected by interference, the noise signal which may significantly affect the Signal to Interference-plus-Noise Ratio (SINR), leading to transmission errors. Moreover, wireless links have their own set of transmission parameters, which have to be properly tuned in order to suit the mesh network to the traffic dynamics. Thanks to the availability of cost-effective radio interfaces, it is possible to endow each wireless node with multiple radios, thus exploiting simultaneous transmissions on the available orthogonal channels and reducing interference. However, a fundamental issue in the design of multi-radio wireless mesh networks is the joint channel assignment and routing problem. The problem consists in how to assign the available channels to the radio interfaces, and how to find the set of flow rates that have to be routed on each link in order to achieve a given objective. The joint problem, though, is NP-complete, and mainly heuristics have been proposed, which solve the two problems separately. Therefore, in this Thesis we solve the routing problem and focus on the channel assignment problem in multi-radio wireless mesh networks, trying to find a channel assignment such that a given set of nominal flow rates can be scheduled. Such problem is shown once again to be NP-complete, hence heuristics are developed. We study the influence of typical wireless parameters, like transmission data rate, transmission power and transmission frequency, on the overall performance of a wireless mesh network and present two different approaches to channel assignment. Assuming that the set of flow rates is given, we introduce a channel, rate and power assignment algorithm that aims at reducing the interference inside the mesh network, while trying to ensure scheduling for the nominal flows. Since the wireless parameters configuration depends on the set of network flow rates, when the traffic changes, then the channel assignment has to be recomputed, i.e., it has to be adapted to the new rates. Therefore, we analyze the issues that arise when channels are re-assigned without taking into consideration the current channel assignment, and present an efficient channel and rate re-assignment algorithm. Once the channel assignment and routing problems are solved, we have the proper channel configuration and the nominal set of flows to be routed on the wireless links. The traffic routing operations, though, are still performed by the network routing agent, which usually route traffic along the minimum hop-count path. In this Thesis, then, we present a routing protocol that is aware of the pre-computed flow rates, and tries to route packets in order to obtain an average link utilization that is as close as possible to the nominal one. We modify a minimum hop-count routing protocol so as to consider variable costs for each link, and we propose a local search heuristic to find the best link cost that allows to optimize the link utilization. Internet evolution has also been accompanied by new protocols, new communication paradigms and new user interactions. Peer-to-Peer applications are an exemplary case of the new trend of Internet, and they create nowadays more than half the Internet traffic. One of the most distinguishing characteristics of Peer-to-Peer applications is that they communicate by means of overlay networks, unaware of the physical network topology they are built upon. Therefore, Peer-to-Peer applications represent a strong engineering challenge for the network management, since the only way to control the traffic is to control the decisions that the applications make at the overlay level. We study the overlay/underlay issues in the wireless mesh networking context, and outline how controlling P2P traffic is even more stringent, due to the mesh configuration sensibility to traffic patterns, and to the restraint of resources. To control the Peer-to-Peer application traffic, we have to somehow guide the overlay applications in the choice of the connection end points, providing information about the physical underlying network. To propose innovative solutions to the overlay traffic optimization problem, the IETF ALTO Working Group has been recently founded. The aim of the ALTO Working Group is to define a structured architecture to support the cooperation between overlay and physical networks. In this context, then, we develop and implement a new component for the ALTO Service architecture, the ALTO Agent, designed to guide the overlay topology creation and to help the Peer-to-Peer application in selecting the best nodes inside a wireless mesh network. In order to choose the best available nodes and to evaluate the performance of wireless paths, we introduce an interference-aware path metric. With the path metric we propose, each link is penalized by taking into account the interference with previous hops on the path. Given the importance of testing our solution in a realistic scenario, we study the integration problem of OMF-based wireless testbeds into the PlanetLab testbed. Thanks to our efforts, we have been able to exploit a realistic large scale testbed to test our ALTO Agent solution. Finally, the ALTO Agent has been installed inside a wireless mesh network, and it has been used to retrieve performance metrics about end-to-end connections between peer mesh nodes. We evaluate the performance of the ALTO Agent into our integrated wireless testbed and show the improvements obtained for the application end user and the reduced costs for the network operator