No abstract

The growing complexity of mobile applications coupled with slow progress in the development of batteries has led to the requirement of energy-awareness in mobile devices. Nevertheless, no general solution exists for supporting energy-awareness across various mobile platforms and application domains. To address the above mentioned problems, we propose a middleware framework which utilizes the concept of application classification, and power estimation to accomplish application-specific power management, as well as providing basic support for active power management and fundamental services for energy-aware applications. To this end, we have implemented a basic prototype reflecting the functionalities of our framework, and evaluated it using mobile YouTube.

Prediction of wireless network conditions enables the reconfiguration of mobile applications in a varying network environment, which in turn might gain more energy savings and better quality of service. In this paper, we focus on the prediction of network signal strength and its potential of improving energy saving in network-based power adaptations. We evaluate the performance of three prediction algorithms, namely, ARIMA, Linear regression and NFI, based on the data sets collected from diverse real-life network environments. Later, we apply the network prediction algorithms to adaptive file download, and compare their effectiveness in terms of energy savings. The results show that the adaptations using prediction could save up to 14.7% more energy when compared to prediction-less adaptation.