| With the rapid growth of the Internet-based businesses and applications that heavily depend on large multimedia contents, the quality of connection to the Internet and the size of available bandwidth have become primary concerns for network operators, ISPs, web site owners, web based application service providers (ASPS), e-business companies, and consumers. The fundamental design of the Internet, however, provides "best-effort" service for all users. What quality of service (QoS) levels to provide and how to allocate traffic among the levels is an urgent research topic and is critical for the continued development of online businesses and the digital economy.;This study focuses on the operationalization issues of differentiated service (Diffserv), which has been recently proposed by computer scientists and engineers aimed at creating the capacity for user-selectable multiple QoS levels through out the Internet. Although Diffserv provides the technical means to differentiate Internet traffic, many economic problems are left unanswered. My dissertation examines several mechanism and policy design issues through economic analysis and computer simulation.;The first part of this dissertation attempts to analyze resource allocation issues in Diffserv. It reviews several currently proposed frameworks, including Paris Metro Pricing (PMP) and leaky-token bucket contracts, and points out the inefficiency introduced by these frameworks with rigorous mathematical analysis. Then this research takes a system-wide optimization approach and finds usage-based pricing is the most efficient pricing scheme. In order to assess the magnitude of comparative difference among different pricing mechanisms, numerical solutions to specific examples are given in the dissertation. Finally, I use simulation to setup a more realistic network environment and study the implementation issues related to the proposed pricing scheme. |
