Resource allocation in differential service networks via network pricing and admission control

Optimal network resource allocation is an important networking research topic, and it is the research problem addressed in this dissertation. In particular, we formulate and study the network pricing- and admission control-based resource allocation schemes, where we view pricing and admission control as two stages of a unified allocation scheme.;We consider multi-traffic-class loss networks where a network provider has the freedom to set a pricing policy and to apply a CAC policy for performance optimization. We study both the loss rate unconstrained and the loss rate constrained problems, where the performance objective of the former is to maximize the network revenue and the performance objective of the latter is to maximize the network revenue subject to the loss rate requirement of each traffic class. Additionally, we apply our resource allocation scheme to the public safety and commercial spectrum sharing problem.;For the loss rate unconstrained problem, we show that, relative to the performance via the arbitrarily chosen prices and the trivial greedy CAC policy, the impact of the greedy optimal pricing is greater than that of an optimal dynamic CAC. We also show that the greedy optimal pricing is a sound pricing policy for practical use, and no additional joint optimization of pricing and CAC is necessary. Moreover, in large networks with many small users, the performance resulting from the greedy optimal pricing and the greedy CAC policy approaches the upper bound performance asymptotically. For the loss rate constrained problem, we demonstrate that both pricing and CAC are useful for performance improvement. Moreover, with the use of the two-stage resource allocation, a higher revenue can be achieved while meeting the call-level loss requirements simultaneously. When applying our network pricing- and admission control-based resource allocation framework to the public safety and commercial spectrum sharing problem, we show that, relative to the segregated networks using the same amount of spectrum resource, our sharing strategy not only satisfies the objective of public safety communication, but also achieves superior commercial revenue performances, at lower prices to consumers.