动态电源与频率调整技术能够帮助实时系统显著减少能耗,之前的研究大多聚焦于基于周期性任务的线程调度算法,却很少考虑周期性与非周期性任务混合的模型.同时,尽管基于CPU利用率的DVS算法可以从系统级上减少能耗,但不能保证实时性.本文提出一种新的算法,它结合减慢因子的DVFS调度算法与系统级的DVS技术,融合PID控制器与自适应的权衡策略为软实时系统提供更好的能耗减少方法.该算法的能耗在服务器利用率低于25％的情况下比加州大学提出的算法下降了14.2％～25.9％,周期性任务超过时限率低于3％.%Dynamic voltage and frequency scaling can significantly save energy for real-time systems.Previous studies consider thread schedules focused on the model of periodic tasks set while only a few of those heuristics aim at the mixed tasks set of periodic tasks and aperiodic tasks.Moreover,DVS heuristics based past value of CPU utilization reduce the energy consumed at the system level,but most of them cannot guarantee real-time constraint.This paper presents a new approach to combine slowdown factor voltage/frequency scheduling and system-level DVS technique with PID controller and adaptive trade-off policy for better energy savings in soft real-time systems.This approach shows that when the utilization of server is less than 25％,the energy dissipation is reduced by 14.2％～25.9％ less than 3％ deadline missed rate compared with the proposed heuristics by University of the California.
展开▼
