| dc.description.abstract | This thesis validates that packet aggregation is a viable technique to increase call capacity for Voice over Internet Protocol over wireless mesh networks. Wireless mesh networks are attractive ways to provide voice services to rural communities. Due to the ad-hoc routing nature of mesh networks, packet loss and delay can reduce voice quality. Even on non-mesh networks, voice quality is reduced by high overhead, associated with the transmission of multiple small packets. Packet aggregation techniques are proven to increase VoIP performance and thus can be deployed in wireless mesh networks. Kernel level packet aggregation was initially implemented and tested on a small mesh network of PCs running Linux, and standard baseline vs. aggregation tests were conducted with a realistic voice traffic profile in hop-to-hop mode. Modifications of the kernel were then transferred to either end of a nine node 'mesh potato' network and those tests were conducted with only the end nodes modified to perform aggregation duties. Packet aggregation increased call capacity expectedly, while quality of service was maintained in both instances, and hop-to-hop aggregation outperformed the end-to-end configuration 4:1. However, implementing hop-to-hop in a scalable fashion is prohibitive, due to the extensive kernel level debugging that must be done to achieve the call capacity increase. Therefore, end-to-end call capacity increase is an acceptable compromise for eventual scalable deployment of voice over wireless mesh networks. | en_US |
