结合Backfilling和空闲资源调度的云工作流调度方法  

A Cloud Workflow Scheduling Scheme Based on Combining Backfilling and Idle Resource Scheduling

作　　者：谭海中[1] 赵丽 TAN Hai-zhong;ZHAO Li(Information Technology Department, Guangzhou Institute of Technology, Guangzhou 510925, China;School of Software, Shanxi University, Taiyuan 030006, China)

机构地区：[1]广州工程技术职业学院信息工程系,广州510925 [2]山西大学软件学院,太原030006

出　　处：《西南大学学报（自然科学版）》2018年第6期149-157,共9页Journal of Southwest University(Natural Science Edition)

基　　金：山西省基础研究计划项目--青年科技研究基金(2014021039-6)

摘　　要：针对云计算中工作流的科学调度问题,提出了一种快速且有效的调度方案.首先,根据计算速度将所有的资源节点以降序方式排列;然后,调度程序通过深度优先搜索,检查任务之间的依赖关系,并根据截止期限对待执行任务进行加权排序;接着,计算每个待执行任务所使用的资源的时隙.如果当前可用资源不能满足当前任务,则采用Backfilling策略,对该任务所需资源进行预留,并跳到下一个任务执行.如果当前资源满足当前任务,则执行提出的空闲资源调度(IRS)策略,尽量安排空闲资源来执行该任务.仿真结果表明:与当前云工作流调度技术相比,本文调度策略具有更低的任务完成时间与任务执行延迟,以及更高的资源利用率.A fast and effective scheduling scheme is proposed in this paper for solving workflow scheduling problems in cloud computing.First,all resource nodes are arranged in a descending order according to the calculation speed.Then,the scheduler evaluates the dependencies between the tasks by depth-first search（DFS）and weights the tasks according to the deadline.After that,the time slot of each resource to be used for the task is calculated.If the currently available resource does not meet the current task,the backfilling policy is used to reserve the resources required for the task and skip to the next task.If the current resource satisfies the current task,the proposed idle resource scheduling（IRS）policy is performed to schedule the idle resource to perform the task.Simulation results show that the proposed scheduling strategy has excellent performance in terms of task completion time,task execution delay and resource utilization.

关 键 词：云计算 工作流调度 Backfilling策略 空闲资源调度 任务执行延迟 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]