Competitive traffic pricing for the internet
Joseph Hui, Prabhanjan C. Gurumohan · 2005
Pricing and consequential revenue maximization by each service provider in the Internet under time varying demand is the focus of this thesis. With competitive prices and differentiated quality of services, users have multiple options to route traffic over the Internet. This dissertation investigates the underlying structure of the interaction between pricing and routing in the Internet in order to aid the revenue maximization objective of service providers and routing cost minimization for users. An analytical model that combines the revenue maximization objective of each service provider with users routing objectives is formulated. The routing objectives are formulated under two different objectives; namely social and selfish. The formulation is used to investigate if service prices are at equilibrium. The formulation also aids in understanding the behavior of competition and/or collusion of service providers. The service prices that maximize each service providers revenue is proved to be at equilibrium. The existence of equilibrium is discovered to be based primarily on the form of service provider's revenue function. It is proved that price equilibrium exists if the revenue functions are bilinear, quasi concave, or concave functions. The conditions for uniqueness of price equilibrium are also developed. A practical algorithm that helps to determine prices in an independent and distributed manner by each service provider is developed. The algorithm is based on Newton's method for solving equality constrained nonlinear optimization problem. Based on whether the routing objective is socially or selfishly optimal, the combined problem of revenue maximization and routing is solved as a bilevel or single level nonlinear optimization problem. Revenues under social and selfish routing are compared. It is proved that prices and revenues are higher under social routing; leading to the conclusion that selfish routing is better in terms of the price paid by the users for service. Unlike the case of social routing, it is shown that service providers can price according to the quality of service offered under selfish routing. Finally, implementation issues of including dynamic service prices and consequential traffic management in the internet are investigated based on constrained based routing under multi protocol label switching. A brief study of regulations and its influence on market structure is also presented.