Since the quality of contents and performance of mobile devices have been dramatically improved, power control technique for mobile devices became an important issue. Most of mobile devices have a disk drive which has large capacity, to store the high...
Since the quality of contents and performance of mobile devices have been dramatically improved, power control technique for mobile devices became an important issue. Most of mobile devices have a disk drive which has large capacity, to store the high quality contents. Since the disk consumes most of the power of a mobile device, we need disk power management technique to improve the performance of mobile device with limited power. Disk power management techniques use the fact that a disk has four different states, active, idle, standby and sleep, and the amount of disk power consumption is different in each state. The disk power management which uses Prefetching is an active and practical method.In this paper, we propose a disk power management technique based on dynamic change of the prefetch buffer size. In this technique, Kernel decides whether the application has characteristics of periodic or interactive applications, and changes the prefetching size based on the characteristics. We analyzed the effect of the prefetching. Prefetching reduce the disk power consumption with periodic applications. To validate this fact, we analyzed the influence of prefetching to executing time of applications and disk power consumption both with periodic applications and interactive applications. Furthermore, we propose a disk power model and optimal Kernel Prefetching size through the model and experimental results. Our method is implemented on Harmonia, a system power management framework, which is based on Linux 2.6.9.