基于智能的分布式算法的设计与实现

发布时间：2018-06-25 22:24

本文选题：智能 + 分布式算法　；参考：《南京邮电大学》2012年硕士论文

【摘要】：随着计算机网络的普遍应用,分布式系统已成为计算机应用的基本形式。由于分布式算法是分布式系统中的关键技术之一,并影响着系统的性能,所以它得到了学者和工程师们的高度关注。在简要介绍了分布式算法和智能算法后,叙述了本文主要研究的任务分配算法和协调者选举算法的研究现状。本文采用智能计算中的粒子群算法解决分布式系统的独立任务调度问题。针对算法过早地陷入局部极值的缺点,提出自适应惯性权重和保持种群多样性的规则;为了保持系统的负载均衡,建立新的适应度函数;将改进的算法应用于任务调度问题,在建立了一个简单网络模型后,对算法进行测试。仿真结果表明,算法在独立任务调度中具有良好的性能,并在负载均衡方面相比其他算法有明显的改善。协调者选举是分布式计算中的一个基本问题。在分析了经典的分布式选举协调者算法后,提出了启发式搜索和消息并行传递的思想;建立一种双向环的选举模型,并据此设计了相应的选举算法;对该算法进行了性能分析,并与经典的选举算法进行比较。结果表明,新的算法在选举出新协调者所花费的时间更少。上述工作证实,将智能的思想引入到分布式算法的设计,能使程序快速、合理的运行,从而提高系统的效率。
[Abstract]:With the widespread application of computer network, distributed system has become the basic form of computer application. Because distributed algorithm is one of the key technologies in distributed system and affects the performance of the system, it has been highly concerned by scholars and engineers. After a brief introduction of distributed algorithm and intelligent algorithm, the research status of task assignment algorithm and coordinator election algorithm, which is mainly studied in this paper, is described. In this paper, particle swarm optimization (PSO) algorithm is used to solve the independent task scheduling problem in distributed systems. In order to maintain the load balance of the system, a new fitness function is established, and the improved algorithm is applied to the task scheduling problem, aiming at the disadvantage that the algorithm falls into local extremum prematurely, and puts forward the rules of adaptive inertial weight and population diversity, in order to maintain the load balance of the system, establish a new fitness function. After establishing a simple network model, the algorithm is tested. The simulation results show that the algorithm has a good performance in independent task scheduling, and has a significant improvement compared with other algorithms in load balancing. Coordinator election is a basic problem in distributed computing. After analyzing the classical distributed election coordinator algorithm, this paper puts forward the idea of heuristic search and message passing in parallel, establishes an election model of bidirectional loop, and designs the corresponding election algorithm accordingly, and analyzes the performance of the algorithm. And compared with the classical election algorithm. The results show that the new algorithm takes less time to select the new coordinator. The above work proves that introducing the idea of intelligence into the design of distributed algorithm can make the program run quickly and reasonably and thus improve the efficiency of the system.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP338.8

【引证文献】