Pricing and efficiency in wireless cellular data networks

Abstract

In this thesis, we address the problem of resource allocation in wireless cellular networks carrying elastic data traffic. A recent approach to the study of large scale engineering systems, such as communication networks, has been to apply fundamental economic principles to understand how resources can be efficiently allocated in a system despite the competing interests and selfish behavior of the users. The most common approach has been to assume that each user behaves selfishly according to a payoff function, which is the difference between his utility derived from the resources he is allocated, and the price charged by the network's manager. The network manager can influence user behavior through the price, and thereby improve the system's efficiency. While extensive analysis along these lines has been carried out for wireline networks (see, for example, [10], [7], [23], [29], [21]), the wireless environment poses a host of unique challenges. Another recent line of research for wireline networks seeks to better understand how the economic realities of data networks can impact the system's efficiency. In particular, authors have considered the case where the network manager sets prices in order to maximize profits rather than achieve efficient resource allocation; see [1] and references therein.

(cont.) In this thesis, we make three contributions. Using a game theoretic framework, we show that rate-based pricing can lead to an efficient allocation of resources in wireless cellular networks carrying elastic traffic. Second, we use the game theoretic equilibrium notions as motivation for a cellular rate control algorithm, and examine its convergence and stability properties. Third, we study the impact of a profit-maximizing price setter on the system's efficiency. In particular, we show the surprising result that for a broad class of utility functions, including logarithmic and linear utilities, the profit maximizing price results in efficiency.