一种基于改进遗传算法的雾计算任务调度策略

摘要

任务的调度与分配一直以来都是云计算技术发展中的关键问题.然而,随着物联网连接设备的爆炸式增长,云计算已不能满足一些任务的调度需求,如健康检测、应急响应等都需要较低的延迟,雾计算应运而生.雾计算将云的服务扩展到网络边缘.雾计算架构下的任务调度与分配目前还是一个较新的研究热点.文中介绍了一种改进的遗传算法(IGA),该算法将适应度判断引入到亲代变异操作中,克服了基本遗传算法(SGA)在变异操作中的盲目性.在雾计算架构下采用该算法调度任务时考虑了服务等级目标(SLO)中响应时间的约束(FOG-SLO-IGA).实验结果表明,FOG-SLO-IGA调度用户任务时在时延、SLO违规率以及服务提供商的花费上均低于云计算架构下采用IGA的调度(CLOUD-IGA);同时,在雾端调度任务时,IGA算法在执行速度上要快于传统SGA算法和轮询调度算法(RRSA).%Task scheduling and assignment has always been a key issue in the development of cloud computing.However,with the explosive growth of internet connection devices,cloud computing has been unable to meet some requirements such as health monitoring,emergency response and so on,which all require low latency.Thus fog computing appears.Fog computing extends the cloud services to the edge of network.Under the fog computing architecture,task scheduling and assignment is still a relatively new research hotspot.This paper introduced an improved genetic algorithm(IGA).The algorithm introduces the fitness judgment into the parental mutation operation which overcomes the blindness of simple genetic algorithm(SGA) in mutation operation.The response time restriction in the service level objective(SLO) is considered when the IGA is used to schedule tasks(FOG-SLO-IGA).The experimental results show that when scheduling user tasks are under the fog computing architecture,FOG-SLO-IGA is superior to the scheduling which uses IGA under cloud computing architecture in latency,SLO violation rate and service provider's cost.Futhermore,IGA algorithm is superior to the traditional SGA algorithm and the round-robin scheduling algorithm(RRSA) in the execution of the tasks under the fog computing architecture.