Packet voice over non-QoS networks.

Abstract

Past communication networks were characterized by separate networks for data and voice. The integration of data and voice in a unique network rapidly gained more and more interest. This integration can provide bandwidth efficiency and new integrated services. The traditional packet switched network cannot be directly applied to voice communication, because it doesn't provide quality of service (QoS). One solution is to modify the network structure to provide new real-time services. Application adaptation provides an alternative solution, with which we can build real-time applications upon currently available network services. The adaptations include bandwidth control, lost packet recovery, delay jitter control, and operating scheduling compensation. The key problem is delay jitter compensation. In this thesis, the application adaptation solution is discussed and algorithms are surveyed. A simulation tool is developed to study the jitter management problem. By managing the tradeoff between display latency and gap frequency, and by dynamically adjusting the display latency according to network conditions. we can provide acceptable voice transmission as long as the average bandwidth is sufficient.