等价类划分的粗粒度任务调度算法  

Coarse-grain Task Scheduling Algorithm For Equivalence Class Partitioning

作　　者：刘东领 袁景凌[1,2] 陈旻骋 

机构地区：[1]武汉理工大学计算机科学与技术学院,武汉430070 [2]交通物联网技术湖北省重点实验室(武汉理工大学),武汉430070

出　　处：《小型微型计算机系统》2018年第2期347-351,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61303029)资助;科技部科技支撑计划子项目(2012BAH89F01)资助

摘　　要：云计算环境下,用户任务请求的多样化及资源的异构性,使任务调度成为云计算中的关键问题.结合任务和资源异构性、多样性的特点,本文提出了一种等价类划分的粗粒度任务调度算法.该算法首先对任务和资源属性进行量化,建立任务和资源模型,然后运用等价类划分思想将任务和资源进行粒度划分,使每一组任务按指令长度分配给能力匹配的资源组,缩小组内单个任务选择资源的范围.此外,在每一组任务粒度调度中使用贪心调度策略,提高调度整体性能.实验结果表明,相比与顺序调度和聚类调度,这种粗粒度任务调度算法在任务总执行时间和完成时间上都具有优越性.In cloud computing environment,diversification of user task requests and resource heterogeneity make task scheduling a key issue in cloud computing.Combined with the heterogeneity and diversity of tasks and resources,a coarse-grain task scheduling algorithm for equivalence class partitioning is proposed.Firstly,quantizing the tasks and resources,and establishing task and resource model.Then,the tasks and resources are partitioned by the equivalence class,so each task group can be allocated to the appropriate resource group by capability,and reduce the scope of a single task to choose resource.In addition,greedy algorithm is used to improve the overall performance of scheduling in each group.The research results show that the coarse-grain task scheduling algorithm has superiority on the total execution time and completion time compared with the order scheduling and clustering scheduling.

关 键 词：云计算 任务调度 粗粒度 等价类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]