Modeling and managing energy consumption of mobile devices

Thanks to the significant improvement in the processing and networking capabilities of mobile devices, mobile devices today can run applications that require complex computation and high network bandwidth. As these applications become ever more popular, a rise is seen in the energy demand that is generated by a typical usage of mobile devices, with the result that existing battery technology is not able to satisfy the growing demand. Improving the energy efficiency of mobile devices and applications has, therefore, become essential. In this thesis, we investigate the energy consumption of mobile devices and propose practical solutions for improving the energy efficiency of wireless data transmission. We propose power models of wireless data transmission over Wi-Fi and show how the power consumption is related to power-saving mechanisms, to Internet traffic characteristics, and to the network throughput. We utilize the linear dependency of transmission costs on network throughput in order to extend the linear regression power models from microprocessor level to system level. These power models provide us with an insight into developing software with energy-efficient wireless data transmission. In this thesis, we present three strategies for reducing transmission cost: applying lossless data compression to network traffic data, scheduling the transmission based on the prediction of network conditions, and power management of the wireless network interface based on the predicted traffic intervals. Our strategies consider the trade-offs between computational and transmission costs, and between energy consumption and transmission performance. In addition, we apply statistical methods for implementing prediction utilities. Finally, considering the complexity in the context collection and processing, we propose an event-driven framework that can be used for implementing, deploying and managing various energy-efficient strategies on mobile platforms