| 802.11-based wireless mesh networks do not scale well with regard to capacity unless the average distance (number of hops) between the source and destination remains small. As the network topology grows larger and traffic increases, channel utilization and throughput decreases. This is due to the 802.11 media access control (MAC) layer failing to achieve an optimal schedule because each node's ability to send is affected by the amount of competition it experiences for the medium. In addition, wireless mesh network operations are hop-centric in that packets are terminated and re-initiated at every hop. This increases end-to-end delay due to processing through the physical, MAC/link, and network layers, re-queuing at the network and MAC layers and re-contention for channel access.; Multimedia requirements of the 1990's drove wired and optical network architects to reconsider the inefficiencies of packet switching and consider long proven methods such as circuit-switching to implement Traffic Engineering in order to reduce end-to-end delay. This resulted in the development of Asynchronous Transfer Mode (ATM) and Multi-Protocol Label Switching (MPLS) technologies. Because both are mature and proven technologies for wired and optical network architectures, much research has been done to apply these methods to wireless mesh networks. But optimal performance improvement eludes designers because of differences between the wired/optical and wireless environments in the provision of non-interfering uni-directional internodal links and lack of a wireless circuit switch. This thesis shows through simulation that these requirements can be met through frequency, space, and/or time multiplexing, hardware design, and the development of resource reservation and path establishment protocols which can support true traffic engineering to wireless mesh networks that provide dramatically improved performance in the form of reduced latency and increased throughput. |
