AbstractsComputer Science

Emerging voice over IP networks dictate a need for efficient network resource management schemes to mitigate the best-effort nature of IP. Traditional quality of service and call admission control mechanisms provide network resource allocation to static voice flows with parameters fixed by a selected voice codec. Frequently, the available bandwidth is not efficiently utilized, which results in a lower-than-possible average voice quality and potential wastage of resources. In this thesis, the author proposes a dynamic voice codec determination scheme that selects codec’s on per-call basis with respect to available resources. The proposed mechanism relies on continuous monitoring of a centrally managed IP network to determine the best voice codec selections for each pair of participating voice gateways. The determination process is modeled as a simple knapsack-like problem to take the fullest advantage of available resources, while maximizing the average voice quality. Overall, the new system is shown to achieve a better average voice quality and more efficient network resource utilization, when compared to traditional application centric QoS and all admission control solutions.