| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 93 === Dynamic voltage scaling (DVS) algorithms save energy by scaling down the processor frequency when the processor is not fully loaded. How to schedule the sporadic, hard real-time tasks with shared resource in a power saving way is still an open problem. Thus, in the thesis, the problem of power aware scheduling for sporadic tasks that share a set of serially reusable, single unit software resources is considered. The goals of this work are that (1) each release of each task should be completed before a well-defined deadline, (2) a resource is serially used by more than one task simultaneously and (3) energy is used minimally. A DVS algorithm, called DVSSR (Dynamic Voltage Scaling for Sporadic Tasks with Shared Resource), is presented to solve the problem. DVSSR offers a power-minimized scheduling algorithm in conjunction with preemptive EDF/DDM scheduling to improve the effectiveness and efficiency of task scheduling.
In the simulation, RMS, a real application, is investigated. In this application, DVSSR and other DVS algorithms are simulated and compared. Our DVS algorithm offers reasonable trade-off between cost and power savings. In RMS application, DVSSR achieves 92.03% average power savings. The properties of resource-sharing sporadic task model are also explored in simulation results.
|
|---|
