边缘计算中基于任务分解的任务分配算法  

Task Scheduling Algorithm Based on Task Decomposition in Edge Computing

作　　者：王磊磊 邓晓衡[1,2] 桂劲松 刘斌[3] 付琨[3] WANG Leilei;DENG Xiaoheng;GUI Jingsong;LIU Bin;FU Kun(School of Computer Science and Engineering,Central South University,Changsha 410075,China;Research Institute of Shenzhen,Central South University,Shenzhen 518071,China;Institute of Aerospace Information,Chinese Academy of Sciences,Beijing 100000,China)

机构地区：[1]中南大学计算机学院,湖南长沙410075 [2]中南大学深圳研究院,广东深圳518071 [3]中国科学院航天信息研究所,北京100000

出　　处：《湖南科技大学学报（自然科学版）》2022年第1期76-84,共9页Journal of Hunan University of Science And Technology：Natural Science Edition

基　　金：国家自然科学基金资助项目(61772553,62172441);湖南省研究生创新型项目资助(2020zzts138,CX20200211)。

摘　　要：大规模任务使得任务服务质量遭受到巨大挑战.边缘计算环境能够为大规模任务处理提供很好的执行模式.针对这一问题,提出了一种基于任务分解的任务调度算法(Task Decomposition based task Allocation Friendly algorithm, TDAF).该算法主要包含2个模块:任务分解模块和任务调度模块.在任务分解模块中,设计了任务关联矩阵以及定义了和任务关联性相关的定义介绍,进而通过任务间的相似度和信息输入输出间的关联性得到任务关联性的计算.以最小化任务关联性设计任务分解目标函数,采用遗传算法对任务进行优化分解.在任务调度模块,通过计算分解后的子任务的优先级确定每一个子任务资源分配的方案.仿真结果表明:TDAF算法的任务执行完成时间和吞吐率性能更优.Large-scale task makes task service quality suffer from a huge challenge. Edge computing environments can provide a good mode of execution for large-scale task processing. To address this problem, we proposes a novel Task Decomposition based task Allocation Friendly algorithm(TDAF). The algorithm mainly consists of two modules: task decomposition module and task scheduling module. In the task decomposition module, we design the task association matrix and define the definition introduction related to the task association, and then obtain the calculation of task association through the similarity between tasks and the association between information input and output. In order to minimize task association, the objective function of task decomposition is designed, and genetic algorithm is used to optimize task decomposition. In the task scheduling module, we determine the resource allocation scheme for each subtask by calculating the priority of the decomposed subtasks. The simulation results show that the TDAF algorithm has better completion time and throughput.

关 键 词：边缘计算 任务分解 任务分配 任务关联性 遗传算法 优先级 

分 类 号：TP305[自动化与计算机技术—计算机系统结构]