Title: "Adaptive Rate Voice over IP"

When: March 23, 2005 10:00am-12:00 noon

Where: Large Commons Room, 5^th Fl. IS Bldg.

Who: Boonchai Ngamwongwattana

Committee:
Dr. Richard A. Thompson, Dissertation Advisor, Department of Information Science & Telecommunications
Dr. Martin B.H. Weiss, Department of Information Science & Telecommunications
Dr. Joseph Kabara, Department of Information Science & Telecommunications
Dr. Stephen M. Walters, Associated Vice President, Telcordia Technologies Inc., Retired
Raul Vera, Chief Technical Officer, TeleContinuity, Inc.

Abstract: VoIP typically uses UDP as its transport protocol which does not provide any congestion control mechanism. At the application level, VoIP simply transmits packets at a constant rate, regardless of the state of the network. Such implementation does not address the issue of dealing with network congestion. When network congestion occurs, VoIP still continues sending packets. This results in even larger packet delay and excessive packet loss. This research focuses on this problem. We propose to develop an adaptive rate VoIP system that has the ability to detect the state of network and adapt the transmission rate accordingly in order to optimize delay and loss performance. We take a comprehensive approach in the study and development of adaptive rate VoIP. An adaptive rate VoIP system is composed of three fundamental components: rate adaptation, network state detection, and control mechanism. We carefully look at each component, identify the problems, and make attempts to overcome them. In the rate adaptation component, we propose an alternative of using packetization as a means for rate adaptation. This approach applies to any constant bitrate speech coder. Hence, the rate adaptation itself does not have an impact on voice quality and can be transparent to the user. In the network state detection component, we address a serious challenge. Since the Internet implicitly enforces the end-to-end principle, the endpoints are expected to operate independently without any assistance from the network. With no synchronized clock between the sender and receiver, the receiver has no way to make one-way end-to-end delay measurements. We propose a novel measurement methodology called Sync & Sense of Periodic Stream (SSPS). SSPS has the ability to virtually synchronize the transmission and reception timings of the VoIP session, enabling measurement of the whole spectrum of delays in VoIP. In addition, SSPS can estimate the available network bandwidth as well as detect any state of the network, from mild to severe congestion or overload. In the control mechanism component, we identify several desirable properties of adaptive rate VoIP that have yet been explored. Adaptive rate VoIP should maintain a smooth transmission rate and avoid unnecessary rate oscillation. Adaptive rate VoIP should ensure fair bandwidth allocation among competing adaptive rate VoIP flows. We propose to develop a new control mechanism and seamlessly integrate the three components. The goal is to develop an adaptive rate VoIP system that has superior performance, in terms of delay and loss performance, as well as the desirable properties.