With the increasing demand for Internet access anywhere, anytime, the Internet and wireless cellular systems will converge to a ubiquitous information transport infrastructure. Future proliferation of hybrid wireless/wired networks depends on its ability to support heterogeneous multimedia applications. Unlike traditional data applications, multimedia applications have stringent delay constraint. How to efficiently support multimedia applications with Quality of Service (QoS) provisioning over hybrid networks is a very challenging task. Since the dominant transport layer protocol over the Internet, TCP, cannot meet the challenge, we are motivated to redesign and re-engineer the transport layer protocol to efficiently support multimedia applications over hybrid networks.

In this dissertation, we propose an Additive Increase and Multiplicative Decrease (AIMD) Protocol for multimedia applications. The proposed protocol inherits the AIMD congestion control mechanism of TCP. Unlike TCP, the sender is not obligated to retransmit lost packets, since multimedia applications prefer timely delivery service to fully reliable services and end-to-end retransmissions may introduce intolerable delay. In addition, instead of TCP’s increase-by-one or decrease-by-half control strategy, a pair of parameters can be flexibly chosen by applications according to their throughput smoothness requirements. By analyzing the competitive behaviors of AIMD-controlled flows and TCP-controlled flows, we derive a necessary and sufficient TCP-friendly condition for the parameter pair, such that AIMD flows can fairly share link bandwidth with coexisting AIMD and TCP flows, independent of the link capacity and the number of coexisting flows.

Then, we develop a novel analytical framework for quantifying the performance of AIMD flows in hybrid networks, in terms of flow throughput, packet loss rate, and delay distribution. Based on the analytical results, the AIMD protocol parameters can be fine-tuned with the consideration of the QoS requirements and the wireless link profile. Since our approach requires only the exchange of parameters among the application, the transport layer protocol and the link layer protocol, it preserves the end-to-end semantics of the transport layer protocol and the layered-structure of the Internet.

Extensive simulations have been conducted to confirm the efficacy of the analysis and demonstrate the feasibility of our approach. It has been shown that the AIMD protocol can appropriately regulate multimedia traffic to efficiently utilize the wireless links and fairly share the Internet reiii sources with coexisting TCP flows in a distributed manner. More importantly, the protocol can provide satisfactory QoS to heterogeneous multimedia applications over hybrid wireless/wired networks.

With satisfactory QoS provisioning, the end-systems have more incentives to voluntarily regulate multimedia traffic with the congestion-controlled AIMD protocol, which is vital for stability, integrity, and continuous growth of next generation networks.