| Level 2 WiFi Mesh Networks appeared in 2003 with the first proposal for standardization of the IEEE 802.11s. Over the years the level 2-approach has shown limitations from its own and then the research was directed to the study of level 3 mesh networks. Then, the idea of multiradio mesh networks emerged; the routers are equipped with two or more radios: one for the access and one or more for the relay. The main objective of this thesis is to study the performance of multiradio WiFi mesh networks on a test bed, based on the use of existing technologies. The experimental approach has led us to develop a platform to control the test bed and automated the execution of experiments. Finally, its use has led us to reflect on a model of abstraction based on the game vocabulary that would make the use of the platform more intuitive and thus make accessible to a wider audience.;The study of multiradio wireless mesh networks demonstrated a high level of performance. Limitations have been identified in the equity management due to WiFi branches and jumps. They have been corrected by the use of a simple level 3 QoS policy. A bandwidth limit of 25 Mbps to the access points has reduced the effects of saturation. Weighted fair queuing has improved the equity of 15%. And finally, tunnels have stabilized traffic and allowed to manage the bandwidth sharing.;A new methodology is proposed to easily set up a test-bed. An abstraction model based on the game vocabulary is developed to open the usage of test-bed to users with no knowledge of UNIX. A network experiment is defined as a game play (experience execution) in which players (nodes) move (traffics) pawns (UNIX tools) on a chessboard (the network test-bed) by following rules (simulation behaviour, e.g. lost rate). To enable remote access, the abstraction model is placed in a Web GUI interface.;Keywords: wireless mesh network, multiradio, performance, fairness, equity, TCP, QoS, test-bed, control platform, centralized, autoexecution, abstraction model, game, intuitive.;The platform is developed in a modular way to automate the execution of experiences on a test-bed. An experience is defined by a script which synchronizes the execution of UNIX commands on the nodes in the test-bed. Data processing is also done by the use of external scripts. This allows the platform to be customized to any research problem. Execution and data processing are controlled by the platform which automatically returns the results to the user. |
