Android Power Management Strategy Research Based On Importance Of Load

As the capability of the emerging smart phones keeps increasing rapidly, the power consumption rises significantly. On the other hand, the battery capacity grows slowly because of the limited material. So the power efficiency of smart phone has become another important metric to evaluate a smart phone. The low-power techniques in the hardware level are limited. Thus, most of manufactures save power in the software level. As the administration of the hardware resources, Operating System has been paid special attention in power saving researches.OSes make power control decision basing on the work load of that device, which is monitored dynamically. But the most widely used smartphone OS, Android, aims at improving the users’experience by increasing the interactive speed. Thus, tasks in Android are classified into different priorities according how they influence the users’ experience. Tasks that have higher priorities, which can immediately affect the user, should be responded at once, and processed as soon as possible. Tasks that have lower priorities, which will not directly affect the user, can be processed in a delay.This thesis proposes the task-importance based power saving method. This method is designed particularly for the interactive OSes to improve the users’ experience.Contributions of this thesis are as following:1. Based on the design principle of interactive systems, we proposed the tasks-importance based power saving framework. This framework classifies tasks cased on the importance of tasks. The more user focus a task, the more significant will the latency reduce the user experience. Thus, the task is more important. OS sets proper performance of devices according the importance of the tasks to achieve the higher power efficiency.2. We realized a task-importance based CPU DVFS policy in Android. We characterize the CPU load and the tasks’ importance based on the number of messages that are processed in the main thread. Then scale the CPU frequency and voltage according to CPU load and tasks’ importance. The proposed method could reduce30%power averagely on our platform.3. We realize a power saving policy for the Wi-Fi network card based on the tasks-importance. The network requests are classified into foreground, background and idle. For the foreground requests, Wi-Fi network card needs to offer its peak performance. While for the background requests, Wi-Fi network could reduce its performance by extending its idle state to save more power.