Addressing the Energy Crisis in Mobile Computing with Developing Power Aware Software

Reducing program power consumption by resource restricted devices has recently become a very active research area.The driving force behind this interest is the wide-spread popularity of portable computers, handheld devices, and cell phones. Consequently, there is an accelerating demand for increased battery life in mobile devices. One way in which we can increase battery life is to improve battery technology and to produce devices that consume less power. Alternately, we can take a software-based approach. For example, many devices and hardware components are designed with multiple levels of operating power. Application management software (compilers, runtime, and operating systems) can adjust these levels using static and dynamic techniques to reduce program power consumption. Alternately, such systems can select to off-load computation from mobile devices to more capable, wall-powered computers.Our focus is on software-based techniques for reducing program power consumption. In this paper, we first articulate the problem and describe the state-of-the-art in energy-aware hardware features that can be exploited by software systems. We then present existing techniques that consider static and dynamic information to guide selection of device power levels. We then examine methods for partitioning computation between a battery-powered device and a wall-powered server at the method, task, and program level. Finally, we describe our current research that incorporates and extends these prior works.