Performance Optimization And Qos In Long Distance Wifi Mesh Networks

Nowadays more than40percent of the world population is still living in ruralareas. How to increase the Internet penetration rate in those areas remains an issue.Conventional Internet access approaches, such as cellular connections, satellite links,are too expensive to be operational in terms of infrastructure construction andmaintenance. LDmesh (Long Distance wireless mesh) networks consist of multiplelong distant WiFi links, which are routed to the Internet gateways located in urbanareas through multi-hops. It has the capability of providing low cost, easymaintenance Internet access for rural inhabitants. In addition, at each village, thecommunication units (each consists of an AP and a mesh router) form the localCSMA/CA mesh network to provide multiple routes to the local LDmesh node.However, the LDmesh network is not perfect and has a couple of issues, such as thelack of high efficient MAC protocol design, end-to-end QoS routing, node energyefficiency, etc.The main applications in LDmesh networks are VOIP and video conferencing,which require stringent end-to-end QoS guarantees in the face of fluctuating loss andavailable bandwidth variations. In view of this, we propose a MAR-basedmultiple-interface DSR protocol with end-to-end QoS support: MQDSR. MQDSRintegrates the MAR model and admission control into the multiple-interface DSR toprovide end-to-end throughput and delay guarantees. Simulation results show that inthe operation of the TDMA-type MAC protocol, MQDSR will be in favor of the highand normal priority classes (e.g. VOIP or video conferencing) to support betterend-to-end QoS than that of the best effort class.LDmesh networks usually use solar energy as nodes are deployed in rural areas.In the dissertation, we at first perform outdoor experiments to understand the link lossand external interference characteristics of LDmesh links and then propose a joint bitrates and transmit power adaptation algorithm. The algorithm is able to achieve areasonable tradeoff between link throughput and energy consumption by dynamicallyadjusting bit rates and transmit power according to the interference fluctuation.Outdoor experiments show that, in comparison to the static bit rates and transmitpower schemes, the proposed algorithm is of great efficiency in energy savings with alittle compromise for link throughput.Each LDmesh node deployed in a village is connected to a corresponding local CSMA/CA mesh network deployed in the same village. The local mesh networkprovides multiple routes to the corresponding LDmesh node for a complete Internetcoverage of the village. Channelization (defined as the act that each node is free tochoose its bit rates, transmit power, channel frequencies and channel widths) is able toimprove the network capacity or fairness. Two transmission capacity estimationmodels are introduced to further reduce the measurement complexity whenperforming flexible channelization. As for the symmetric carrier sensing, we at firstexperimentally investigate the carrier sensing range and the carrier sensingrelationship among nodes and then extend a transmission capacity model. A6-node802.11a testbed is used for performance evaluations, which show there is less than10%error in prediction for more than80%cases. For the asymmetric carrier sensing, wepropose a probabilistic model for nodes’ transmission capacity estimation showing theprediction error is less than8%by test-bed experiments.