TABLE OF CONTENTS In this section we give a survey of the current research directions concerning agent and active network techniques applied to adaptive applications. We start with an overview of current techniques used in network and transport level adaptation protocols, and then discuss the potential of mobile agents and active networks for such applications, with a survey of current proposals in this area. After that we review market-based control research applied to resource sharing in computer and telecommunication networks. The use of such techniques in the active network context is the focus of our work.
Adaptive applications can tolerate fluctuations in resource availability, and are necessary in a heterogeneous environment such as the Internet today, where different network technologies and user terminals are interconnected together, and over which a multitude of services coexist. In the case of multimedia applications, a good survey can be found in [VAN 00].
The adaptation mechanisms can be implemented at several layers of the protocol stack, ranging from pure application layer techniques to network level protocols. For example, we can adapt to the available bandwidth using elastic traffic that reduces the data rate generated in presence of network congestion. Fluctuations in delay can be dealt with by using elastic buffers to adjust the play-out time. To deal with CPU and memory bottlenecks, some interaction with the operating system is necessary (see Section 5 of [VAN 00] for examples). Our focus in this paper is on network and transport level mechanisms for adaptation.
