边缘计算环境下基于相关性的任务分区实时低功耗调度算法  

Real-time Low-power Scheduling Algorithm Based on Correlation for Task Partitioning in Edge Computing Environments

作　　者：刘芳[1,3] 陈子煜 马昆 彭敏 何炎祥[1] 胡威[2] LIU Fang;CHEN Ziyu;MA Kun;PENG Min;HE Yanxiang;HU Wei(School of Computer Science,Wuhan University,Wuhan 430072,China;School of Computer Science,Wuhan University of Science and Technology,Wuhan 430081,China;School of Artificial Intelligence,Wuhan Vocational College of Software and Engineering,Wuhan 430205,China)

机构地区：[1]武汉大学计算机学院,武汉430072 [2]武汉科技大学计算机科学与技术学院,武汉430081 [3]武汉软件工程职业学院人工智能学院,武汉430205

出　　处：《小型微型计算机系统》2025年第2期289-296,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(62072346)资助;国家自然科学基金项目(61972293)资助;科技创新2030-“新一代人工智能”重大项目(2021ZD0113304)资助。

摘　　要：在嵌入式实时系统中,边缘智能技术显著提升了计算性能.然而,确保任务时效性、提高效率、降低能耗和系统阻塞仍然是关键研究领域.本研究专注于同质多核系统的任务调度问题,提出了一种名为“基于相关性的任务分区节能调度策略”(CBTP)的节能调度策略.CBTP通过深度分析任务之间的依赖关系,为它们分配最优处理器,减少资源争用和阻塞.为了实现高效的并发访问,采用了多处理器堆栈资源协议(MSRP)和高性能的分区最早截止优先调度算法(P-EDF).同时,CBTP引入了双速节能机制,结合动态电压和频率调整(DVFS)来灵活调整任务的执行速度.实验结果表明,CBTP策略明显优于传统方法,显著降低了系统阻塞和能耗,验证了其在同质多核系统中的卓越性和有效性.这项研究提供了一种新的视角,旨在边缘计算环境下来提升实时系统的调度性能,同时提高调度的能源效率.In embedded real-time systems,edge intelligence technologies have significantly enhanced computational performance.However,ensuring task timeliness,improving efficiency,reducing energy consumption,and mitigating system bottlenecks remain critical research areas.This study focuses on the task scheduling problem in homogeneous multi-core systems and proposes an energy-efficient scheduling strategy called"Correlation-Based Task Partitioning"(CBTP).CBTP allocates optimal processors to tasks by deeply analyzing the dependencies between them,reducing resource contention and blocking.To achieve efficient concurrent access,it employs the Multiple Stack Resource Protocol(MSRP)and a high-performance Partitioned Earliest Deadline First(P-EDF)scheduling algorithm.Additionally,CBTP introduces a dual-speed energy-saving mechanism,combining dynamic voltage and frequency scaling(DVFS)to flexibly adjust task execution speeds.Experimental results demonstrate that the CBTP strategy outperforms traditional methods,significantly reducing system bottlenecks and energy consumption,confirming its excellence and effectiveness in homogeneous multi-core systems.This research provides a new perspective aimed at enhancing the scheduling performance of real-time systems in edge computing environments while improving scheduling energy efficiency.

关 键 词：边缘智能 任务时效性 任务调度 能耗 多核系统 CBTP策略 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]