面向边缘计算平台的半线上任务动态调度方法  

Dynamic semi-online task scheduling method for the edge computing platform

作　　者：赵辉[1,2] 冯南之 王泉 万波[1,2] 王静 ZHAO Hui;FENG Nanzhi;WANG Quan;WAN Bo;WANG Jing(School of Computer Science and Technology,Xidian University,Xi’an 710071,China;Key Laboratory of Smart Human-Computer Interaction and Wearable Technology of Shaanxi Province,Xi’an 710071,China)

机构地区：[1]西安电子科技大学计算机科学与技术学院,陕西西安710071 [2]陕西省智能人机交互与可穿戴技术重点实验室,陕西西安710071

出　　处：《西安电子科技大学学报》2021年第6期8-15,共8页Journal of Xidian University

基　　金：国家自然科学基金(61972302);陕西省重点研发计划项目(2021GY-014,2021GY-086);中央高校基本科研业务费专项资金(JB210309,JB210312)。

摘　　要：当边缘计算平台中存在已知和未知性能的计算节点时,这一场景下的任务调度称为半线上任务调度。由于未知性能节点的影响,一般任务调度方法可能导致任务执行时间或传输时间过长,使得边缘计算平台高能耗的问题更加突出。针对此问题,以能耗优化为目标,提出了一种面向边缘计算平台的半线上任务动态调度方法。首先,考虑边缘计算平台中能耗的主要影响因素,从边缘节点的处理速度、路由延迟和队列长度三个角度,引入边缘节点的任务执行能耗、任务传输能耗和空闲能耗,建立了面向能耗优化的边缘计算平台任务调度模型;其次,对于边缘计算平台中的未知性能节点,先将其性能假设为某个已知节点,形成未知-已知节点之间的映射关系,再不断感知映射双方的任务队列长度来动态调整映射关系,充分利用已有先验知识,提出了一种基于动态映射的半线上任务调度算法,实现能耗优化;最后,在CloudSim平台完成对比试验。实验结果表明,所提方法相较其他方法能有效地降低边缘计算平台的能耗。When there are known and unknown computing nodes in the edge computing platform,the task scheduling in this scene is called semi-online task scheduling.Due to the influence of unknown nodes,the normal task scheduling method may lead to a long makespan or transmission time,which aggravates the problem of high energy consumption on the edge computing platform.To solve this problem,this paper proposes a Dynamic semi-online task Scheduling Strategy(DSS)for the edge computing platform,aiming at energy consumption optimization.First,by considering the main factors affecting the energy consumption on the edge computing platform,the energy consumption of task execution,task transmission and idle are introduced from the perspectives of the processing speed,routing delay and queue delay of the edge nodes,and then an energy consumption optimization-oriented task scheduling model is established.Second,for the unknown node,this paper proposes a dynamic mapping-based semi-online task scheduling algorithm which assumes that the performances of unknown nodes are equal to a certain given node to form the mapping between unknown and known nodes.Then this algorithm dynamically adjusts the mapping relation of both sides through continuous perception of their task queue lengths,thus making full use of prior knowledge and reducing the energy consumption.Finally,a comparative evaluation is performed on the CloudSim platform,with the results showing that the proposed method can effectively reduce the energy consumption on the edge computing platform.

关 键 词：边缘计算 任务调度 未知计算节点 动态映射 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]