使用遗传算法改进的两阶段云任务调度算法研究  

Improved Two Period Cloud Task Scheduling Algorithm with Genetic Algorithm

作　　者：李腾耀 张凤琴[1] 王梦非[1] 

机构地区：[1]空军工程大学信息与导航学院,西安710077

出　　处：《小型微型计算机系统》2017年第6期1305-1310,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(71503260)资助;陕西省自然科学基金项目(2014JM8345)资助

摘　　要：为了解决传统整数规划方法在云资源调度问题上收敛速度慢,难以适应大规模云端任务调度优化的缺陷,基于遗传算法提出了初始任务配置算法和动态任务配置算法,分别用于解决云端任务初始提交阶段和任务动态运行阶段的资源调度优化问题.在两阶段任务调度优化过程中,分别结合截止时间和资源利用率确定了有针对性的优先级队列,分别使用滑动窗口机制和在线迁移机制提升任务调度性能.通过对迭代过程和收敛速度的实验分析,本文算法能够利用遗传算法的优势解决两阶段云任务调度优化问题,并具有更快的收敛速度.To solve the low convergence speed and shortage by traditional integer planning method on large scale task scheduling, initial task configuration algorithm and dynamic task configuration algorithm was proposed based on genetic algorithm. And cloud resource scheduling problem with initialized accepted period and dynamic running period for tasks could be solved with the two algo- rithms. During the two periods, priority queues were constructed with deadlines and resource ratios respectively. At the same time, slide window mechanism and online migration mechanism were used to improve scheduling efficiency. By analyzing the iterated procession and convergence speed with experiments, cloud resource scheduling with two periods can be solved with higher convergence speed.

关 键 词：遗传算法 资源调度 云计算 滑动窗口 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]