Revenue In Resource Allocation Games And Applications

Abstract

This dissertation studies a general class of resource allocation games in computer systems. The applications of these games include sharing network bandwidth, scheduling jobs in data centers and distributing click-through resources in sponsored search. The main focus of the dissertation is the revenue that can be obtained by providers. We investigate the revenue of proportional sharing under a symmetry condition among users, and show how to modify this mechanism to get a competitive revenue without the symmetry condition. We study the weighted proportional sharing mechanism as a natural extension of fair sharing to capture the incentives of revenue maximizing providers.