DVS系统硬实时周期任务动态调度算法  

Dynamic Scheduling Arithmetics for Hard Real-Time Period Tasks in Dynamic Voltage Scaling System

作　　者：吴琦[1] 熊光泽[1] 廖勇[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都610054

出　　处：《电子科技大学学报》2007年第5期842-845,共4页Journal of University of Electronic Science and Technology of China

基　　金：国家高技术研究发展计划;国家863计划资助项目(2003AA1Z2210)

摘　　要：与实时任务的可调度分析不同,实时DVS调度在保证任务截止时间限制同时,还要关注任务执行的处理器功耗。功耗研究一段时间的累积效果,传统基于最坏执行时间的任务调度模型不能满足实时DVS调度需要。该文采用实际执行时间(AET)概率分布建立实时任务调度模型,利用随机分析的方法分析AET分布对实时DVS调度算法性能的影响,提出了基于平均执行时间的实时DVS动态调度算法。实验结果表明,该算法在任务具有不同AET分布情况下可保持稳定的功耗性能。Dynamic Voltage Scaling （DVS） has become a promising method for embedded real-time systems to reduce their power consumption. Unlike schedulability, energy consumption focus on the accumulative effect in a period of time. The traditional real-time scheduling model based on the worst execution time is not satisfied for the need of energy consumption analysis. In this paper the real-time task scheduling model is built according to the probability distribution of actual execution time, and the effect of the Actual Execute Time （AET） distribution on the energy consumption performance of real-time DVS scheduling algorithm is analyzed using stochastic method. The real-time DVS dynamic scheduling algorithm based on the average execution time is presented. The experimental results show that algorithm presented in this paper can hold stable power consumption performance under different AET distribution.

关 键 词：动态电压调整 感功计算 功耗优化 实时系统 任务调度 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]