Optimal resource allocation: Issues and applications

The common theme of the various chapters of the thesis is the issue of optimal resource allocation in information networks with quality of service (QoS) and performance requirements. In particular, the following instances of such problems are studied: (i) Server allocation in an interconnected queueing system which models parallel multi-resolution image formation in automatic target recognition (ATR) and multi-stage inspection in wafer manufacturing; (ii) connection admission control (CAC) in multi-service single-hop wireless networks; (iii) an optimal routing strategy in ad hoc wireless networks with a power consumption performance criterion, and its sensitivity to estimation error in the channel model. In this dissertation, a systematic and analytic study of these problems is developed. Such a study of any system consists of three distinct stages of modeling, mathematical analysis, and solution-evaluation/model-validation. Though each stage reveals specific and distinct information about the optimal solution, only taken together does a complete picture emerge. Hence, associated with each of the above three problems, elements of modeling, mathematical analysis, and solution-evaluation/model-validation are investigated.; The contributions of the dissertation are: (1) Study and analysis of the performance of server allocation policies in interconnected queues with the expected makespan criterion. (2) A systematic approach to connection admission control and service provisioning in a multi-service single-hop cellular network with QoS requirements. The approach provides an appropriate framework for understanding the interaction of physical layer concerns and network layer issues. The approach is illustrated by the solution of the CAC problem in a multi-service single-hop wireless network with outage-based QoS. (3) Sensitivity analysis for an optimal routing strategy in a wireless ad hoc network with distributed information, in the presence of error in channel quality estimation.