云计算环境下资源分配算法的研究

发布时间：2019-02-28 07:30

【摘要】：随着计算机技术的革新与互联网的飞速发展，云计算应运而生。云计算是一种新兴的商业计算模式，它利用成熟的虚拟化技术将大量的基础设施资源集中起来，实现了数据中心资源的按需服务。在云计算中，由于资源具有动态性、异构性、大规模性等特点，如何根据云计算的实际特点制定合适的资源分配策略是目前急需解决的问题。智能优化算法由于其高度并行、自组织、自适应等特性，已经被广泛用于解决云计算的资源分配问题，本文通过研究云计算下的资源分配问题，对现有的资源分配算法存在的问题进行了分析，主要进行了以下方面的研究工作： ①提出一种粒子群结合遗传算法（PSO-GA）的云计算资源分配算法。传统的的粒子群算法、遗传算法在云计算资源分配过程中均容易陷入早熟收敛的缺陷，不能很好解决云计算下的资源分配。针对这一问题，提出PSO-GA资源分配算法，该算法在遗传算法的基础上通过引入种群分割、种群覆盖的概念，并且将粒子群算法中的变异算子应用到PSO-GA算法的变异过程中。实验表明，PSO-GA算法能够有效解决单一的遗传算法和粒子群算法的早熟收敛的缺陷，提高最优解收敛速度和算法执行效率。 ②提出一种改进型人工鱼群算法（IAFA）的云计算资源分配算法。在云计算资源分配过程中，在种群规模较大的情况下，PSO-GA算法收敛速度较慢，不能快速得到全局最优解。为了解决这一问题，本文提出一种改进型人工鱼群算法（IAFA），在原来行为的基础上淘汰了随机行为，增加了跳跃行为，促使了陷入局部最优的人工鱼跳出局部极值继续搜索全局最优；引入生存周期和生存指数的概念，，节约了储存空间，提高了算法的效率。实验表明，IAFA算法能够在种群规模较大的情况下快速收敛并得到全局最优解。 ③扩展了云计算仿真模拟平台CloudSim，对上文提出的算法进行仿真模拟。本文分析和研究了CloudSim的资源分配机制，对CloudSim平台进行重编译，在CloudSim上实现了PSO-GA、IAFA等算法的仿真程序，并对算法进行了模拟验证和对比分析，实验证明了上述两种改进算法的有效性。
[Abstract]:With the innovation of computer technology and the rapid development of the Internet, cloud computing emerges as the times require. Cloud computing is a new business computing model, which uses mature virtualization technology to centralize a large number of infrastructure resources and realize on-demand service of data center resources. In cloud computing, due to the dynamic, heterogeneous, large-scale characteristics of resources, how to formulate appropriate resource allocation strategy according to the actual characteristics of cloud computing is an urgent problem to be solved at present. Intelligent optimization algorithm has been widely used to solve the resource allocation problem of cloud computing because of its highly parallel, self-organizing, adaptive and other characteristics. This paper studies the resource allocation problem in cloud computing. The existing problems of resource allocation algorithms are analyzed, and the main work is as follows: (1) A cloud computing resource allocation algorithm based on particle swarm optimization (PSO-GA) is proposed. Traditional particle swarm optimization (PSO) and genetic algorithm (GA) are prone to fall into premature convergence in the process of resource allocation in cloud computing, and can not solve the problem of resource allocation in cloud computing. In order to solve this problem, the PSO-GA resource allocation algorithm is proposed. Based on the genetic algorithm, the concept of population segmentation and population coverage is introduced, and the mutation operator in particle swarm optimization is applied to the mutation process of PSO-GA algorithm. Experiments show that PSO-GA algorithm can effectively solve the shortcomings of premature convergence of single genetic algorithm and particle swarm optimization algorithm, improve the convergence rate of the optimal solution and the efficiency of the algorithm. In this paper, an improved artificial fish swarm algorithm (IAFA) for cloud computing resource allocation is proposed. In the process of resource allocation in cloud computing, when the population size is large, the convergence rate of PSO-GA algorithm is slow, and the global optimal solution can not be obtained quickly. In order to solve this problem, an improved artificial fish swarm algorithm (IAFA),) is proposed in this paper, which eliminates random behavior and increases jump behavior on the basis of the original behavior. The artificial fish trapped in the local optimum jump out of the local extremum and continue to search for the global optimal; By introducing the concepts of life cycle and survival index, the storage space is saved and the efficiency of the algorithm is improved. The experimental results show that the IAFA algorithm can converge rapidly and obtain the global optimal solution when the population size is large. 3 extend the cloud computing simulation platform CloudSim, to simulate the algorithm proposed above. This paper analyzes and studies the resource allocation mechanism of CloudSim, recompiles the CloudSim platform, implements the simulation program of PSO-GA,IAFA and other algorithms on CloudSim, and makes simulation verification and comparative analysis of the algorithms. Experimental results show the effectiveness of the two improved algorithms.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP18;TP393.09

【参考文献】