基于粒子群和帝国竞争混合算法的云计算任务调度策略研究

发布时间：2018-06-30 01:30

本文选题：云计算 + 任务调度　；参考：《广西师范大学》2017年硕士论文

【摘要】：云计算使用虚拟化技术将各种计算、存储、网络宽带等实体资源整合成一个共享的云服务资源池,再通过任务调度算法为用户提交的任务分配资源。任务调度算法是云计算中最重要的技术之一,效率优良的任务调度算法能够保证云计算平台稳定高效地运行,可以缩短任务的完成时间,减少用户使用云服务的成本,并且能够保证云计算服务提供商的经济收益。云计算环境非常复杂,传统的调度算法已经无法满足云计算任务调度的需求。新兴的智能调度算法在一定程度上提升了任务调度的性能,但是还不够成熟和稳定,在收敛精度和稳定性等方面还存在缺陷。本文的研究动机是通过分析云计算和任务调度算法的关键技术和特点,了解和掌握现有任务调度的模型和算法,然后设计出性能更加良好的调度算法来解决云计算任务调度面临的问题。主要工作如下:(1)介绍云计算、云计算任务调度、云计算任务调度算法的基本理论之后,接着详细分析粒子群算法和帝国竞争算法的基本原理、数学模型,分析这两种算法的优缺点,并对这两种算法的发展和已有的改进进行总结。(2)通过对比分析粒子群和帝国竞争算法的特点,针对帝国竞争算法中殖民地无自主学习能力、不能记录历史最优信息的缺点以及粒子群算法收敛过快的缺点,提出将粒子群和帝国竞争混合的算法,使具有生物启发性的粒子群算法和具有社会启发性的帝国竞争算法融合在一起,达到优势互补的效果;针对帝国算法中殖民地缺乏有效的控制机制调整移动距离和角度大小的缺点,融入粒子群算法的思想使殖民地具有粒子的特性之后,对惯性权重进行自适应调整。(3)将粒子群和帝国竞争的混合算法应用于云计算任务调度,设计编码形式和适应度函数,然后在云计算仿真平台Cloudsim上进行实验,并将实验结果和改进前的算法进行分析对比。实验结果表明本文算法具有更加良好的性能。
[Abstract]:Cloud computing uses virtualization technology to integrate various computing, storage, network broadband and other physical resources into a shared pool of cloud services resources, and then assign resources to tasks submitted by users through task scheduling algorithms. Task scheduling algorithm is one of the most important technologies in cloud computing. The efficient task scheduling algorithm can ensure the cloud computing platform to run stably and efficiently, can shorten the completion time of tasks and reduce the cost of using cloud services. And can guarantee the economic benefit of cloud computing service provider. Cloud computing environment is very complex, the traditional scheduling algorithm can not meet the needs of cloud computing task scheduling. The new intelligent scheduling algorithm improves the performance of task scheduling to a certain extent, but it is not mature and stable, and there are still some defects in convergence accuracy and stability. The motivation of this paper is to understand and master the existing task scheduling models and algorithms by analyzing the key technologies and characteristics of cloud computing and task scheduling algorithms. Then we design a better scheduling algorithm to solve the problem of cloud computing task scheduling. The main work is as follows: (1) after introducing the basic theory of cloud computing and cloud computing task scheduling algorithm, the basic principle and mathematical model of particle swarm optimization algorithm and imperial competition algorithm are analyzed in detail. The advantages and disadvantages of these two algorithms are analyzed, and the development and improvement of these two algorithms are summarized. (2) by comparing and analyzing the characteristics of particle swarm optimization and imperial competition algorithm, the colony has no autonomous learning ability in imperial competition algorithm. Because the historical optimal information can not be recorded and the convergence of particle swarm optimization algorithm is too fast, a hybrid algorithm of particle swarm optimization and imperial competition is proposed. The particle swarm optimization algorithm with biological enlightenment and the imperial competition algorithm with social enlightenment are combined to achieve the effect of complementary advantages; In view of the lack of effective control mechanism to adjust the distance and angle of the colony in the Imperial algorithm, the particle swarm optimization (PSO) algorithm is used to make the colony have the characteristics of particles. The inertia weight is adjusted adaptively. (3) the hybrid algorithm of particle swarm optimization and imperial competition is applied to the task scheduling of cloud computing, the coding form and fitness function are designed, and then the experiment is carried out on Cloudsim, a cloud computing simulation platform. The experimental results are compared with the improved algorithm. Experimental results show that the proposed algorithm has better performance.
【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP3

【参考文献】