Seamless support for long-lived multi-gigabit flows in wireless networks

Wireless networks in the near future must be able to support seamless mobility for multi-gigabit flows due to growing popularity of smart mobile devices and due to growing demand for applications such as media streaming, video conferencing, and high-definition television. The challenge, however, is that factors such as packet losses, handoff delays, and re-routing delays have more severe performance impact when the data rate increases. Mesh networking is a promising approach to support such high data rates to mobile users in malls, homes, business enterprises, public transport, and personal vehicles. However, to support these multi-gigabit data rates, the wireless mesh networks need access to a large amount of frequency spectrum. Unfortunately, in the current static licensing approach used by spectrum regulatory bodies, unfettered access to a large amount of frequency spectrum is not possible. As a result, spectrum regulatory bodies such as Federal Communication Commission (FCC) are considering a new dynamic frequency allocation approach enabled by recent technological advances called cognitive radios. This dissertation addresses the challenges of multi-gigabit communication in cognitive radio based wireless mesh networks. To address these challenges, the dissertation envisions a network in which multiple service provides deploy their own infrastructure and compete with each other to provide gigabit communication services to mobile users. In order to compete with others, a service provider deploys a set of wireless access points which communicate with each other to form a mesh network over the geographic area of interest. The transmission power of each such access point determines its coverage area such that mobile users within this area can potentially associate with it. Larger the transmission power, the larger coverage area and hence potentially more users. However, since the mobile users associated to an access point must share its wireless bandwidth, more users often results in lower link layer throughput for each user. As a result, mobile users selfishly switch their association between access points, both within the same service provider and also across service providers, in order to increase their individual link layer throughput. This dissertation formulates the problem of configuring the transmission powers of access points of a single provider as a cooperative game between the access points and between the mobile users and the access points of a provider. The game results in a joint optimization of the transmission power selection and user association. The dissertation describes a solution to this optimization problem which in turn results in a method for a single service provider to deploy a wireless mesh network covering the entire geographic area of interest and thereby providing basic wireless access service to the mobile users. In order to further increase the data rate provided to a mobile user, the service providers lease frequency spectrum from nearby licensed users and resell the spectrum to mobile users. Since larger spectrum means higher data rate but at larger cost, mobile users individually optimize their spectrum purchasing strategy based on the needs of their application and the prices set by the service providers for reselling their individual spectrum. This dissertation proposes a dynamic spectrum pricing strategy for the service providers and a dynamic spectrum purchase strategy for the mobile user in order to collectively increase the data rate supported to the users while individually increasing the revenue for each service provider. Unfortunately, such a spectrum leasing and purchasing strategy results in considerable fluctuations in the wireless data rates within the mesh network of a single provider. Simulation results presented in the dissertation show that these fluctuations cause a disproportionate degradation in the end-to-end performance of Transmission Control Protocol (TCP) based traffic flows to mobile users as data rates increase multi-megabit to multi-gigabit. This dissertation also proposes a novel approach called Spare-bandwidth Rate-based Network Coding (SRNC) to alleviate the performance degradation problems caused by link bandwidth fluctuations. Simulation results presented in the dissertation show that, using SRNC, a single service provider can support multi-gigabit end-to-end data rates to mobile users despite the bandwidth fluctuations caused by the leasing and re-leasing of frequency spectrum from nearby licensed users. Consequently, an integrated approach involving SRNC, joint power and user association strategy, dynamic spectrum pricing strategy, and dynamic spectrum purchase strategy results in a seamless support for multi-gigabit data rates to mobile users in cognitive radio based wireless networks.