Dynamic-oriented Resource Management for Mobile Wireless Systems
The ever-increasing adoption and usage of mobile devices have created explosive demand for wireless data services. In spite of the great efforts made to expand the capabilities of wireless infrastructures and mobile devices, performance issues caused by resource contention keep arising. This calls for optimized resource management solutions which not only efficiently utilize network and device resources, but consistently optimize the performance under various scenarios. The problem is highly challenging considering the inherent resource sharing feature of wireless medium and the unpredictable dynamics in wireless environments. In this dissertation, we present a two-tier dynamic-oriented resource management solution for wireless mobile systems, which involves several novel techniques corresponding to different layers of the network architecture. The first tier is a scheme designed to optimize network infrastructure of wireless networks. We study the cross-layer resource management problem which covers multiple layers from the transport layer to the medium access layer. We present two techniques that are operated on different time scales. One technique relies on Capacity Space Projection to form a virtual capacity space, which contains both the slow time scale and the fast time scale capacity signals. Then the cross-layer resource allocation becomes an optimization problem over the virtual capacity space. Using the other technique, we can investigate the risk imposed by unpredictable fast-time scale and perform risk-benefit analysis by introducing a penalty function into the optimization framework. The second tier is an application layer management scheme applied to the ecosystem consisting of mobile terminals and remote computational resources. This scheme aims to improve user experience by minimizing the energy consumption during a task execution. This is achieved by a mobile application offload system that intelligently migrates the computation-intensive tasks from a mobile application to the back-end server. We present energy profiling models and a runtime decision solver to enable cost-effective offloading in dynamic environments. The two tiers employ distinct resource management strategies, however, they are closely related in the layered architecture of mobile wireless systems.