云计算中任务调度算法的优化与研究

发布时间：2018-05-18 13:06

本文选题：云计算 + 架构　；参考：《兰州理工大学》2014年硕士论文

【摘要】：随着网格计算、普适计算以及计算机通讯技术的快速发展,人们越来越希望能把资源、软件及应用更好地整合在一起,并以服务的形式向外提供给用户,因此云计算应运而生。为了实现资源和服务的整合,需要一个更为通用和面向服务的云架构；同时由于云计算环境的异构性、分布式、自治性以及服务的多样性特征,对云平台调度机制也提出了更高的要求,因此关于云架构及其调度机制的研究得到了业界越来越多的关注。本文首先对云计算的基本理论、云架构和云任务调度机制进行了深入的分析和研究,并总结了当前云架构及调度机制存在的问题。针对这些问题并结合云计算的特点,主要做了以下两个方面的改进和研究： (1)针对云计算环境中的大型复杂任务请求提出了改进的MapReduce模型。改进后的模型在Map过程开始之前,先把大型复杂任务请求转化为DAG图,然后将其转化为最小生成树,简化了复杂任务的执行过程,减少了任务执行时间,有效地提高了任务的执行效率。改进后的模型增强了Hadoop架构的性能,扩大了Hadoop架构的应用领域,使得Hadoop架构不仅可以应用于交互式应用,而且还可以应用于科学计算领域。 (2)针对在云计算环境中利用基本蚁群算法进行任务调度时存在的缺点,对蚁群算法进行了改进,提出了基于兄弟蚂蚁和剩余生命信息素的BBPA算法,然后在CloudSim平台上对BBPA算法进行了仿真模拟,仿真结果表明,改进后的蚁群算法在云计算环境中具有更好的任务搜索效率和任务执行效率最后,对该论文中提出的调度模型和调度算法的创新性方面进行了总结,并对今后云计算的发展和任务调度的研究方向进行了展望。
[Abstract]:With the rapid development of grid computing, pervasive computing and computer communication technology, more and more people hope to integrate resources, software and applications together better and provide them to users in the form of services. Therefore, cloud computing emerges as the times require. In order to integrate resources and services, we need a more general and service-oriented cloud architecture, and because of the heterogeneity, distribution, autonomy and diversity of services in the cloud computing environment, Therefore, more and more attention has been paid to the research of cloud architecture and its scheduling mechanism. In this paper, the basic theory of cloud computing, cloud architecture and cloud task scheduling mechanism are analyzed and studied deeply, and the existing problems of cloud architecture and scheduling mechanism are summarized. In view of these problems and combined with the characteristics of cloud computing, we mainly do the following two aspects of improvement and research: 1) an improved MapReduce model is proposed for large complex task requests in cloud computing environment. Before the process of Map, the improved model transforms the large complex task request into DAG graph, and then transforms it into the minimum spanning tree, which simplifies the execution process of complex task and reduces the task execution time. The efficiency of task execution is improved effectively. The improved model enhances the performance of Hadoop architecture and expands the application field of Hadoop architecture. Hadoop architecture can be used not only in interactive applications but also in scientific computing. 2) aiming at the shortcomings of using basic ant colony algorithm to schedule tasks in cloud computing environment, this paper improves the ant colony algorithm, and proposes a BBPA algorithm based on sibling ants and residual life pheromones. Then the BBPA algorithm is simulated on CloudSim platform. The simulation results show that the improved ant colony algorithm has better task search efficiency and task execution efficiency in cloud computing environment. Finally, the innovative aspects of scheduling model and scheduling algorithm proposed in this paper are summarized, and the future development of cloud computing and the research direction of task scheduling are prospected.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01

【参考文献】