一种基于路径层面的遗传算法  

A path level-based genetic algorithm

作　　者：朱国晖 牛皎月 王丹妮 ZHU Guohui;NIU Jiaoyue;WANG Danni(School of Communications and Information Engineering,Xi'an University of Posts and Telecommunications,Xi'an 710121,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2022年第2期25-31,共7页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金项目(61371087)。

摘　　要：针对当前数据中心网络在路径选择过程中出现的资源分配不合理和资源利用率低的问题,提出一种基于路径层面的遗传算法(Path Level-based Genetic Algorithm,PLGA),以提高网络传输性能。利用软件定义网络(Software Define Network,SDN)在产生遗传算法的初始种群时,引用Yen算法产生第一代种群,对初始种群进行初步优化,选择出前K条最短路径。使用路径可用带宽质量、链路均衡度以及路由跳数等3个指标设计新的适应度函数,再对初始种群进行选择、交叉和变异操作,最终寻找最优路径。实验结果表明,PLGA算法在平均带宽利用率、吞吐量和时延抖动方面的性能优于等价多路径(Equal-Cost Multi-Path,ECMP)、基于多路径传输的动态负载均衡路由(Multipath Transmission-based Dynamic Load-balanced Routing,MTDLR)和全局负载均衡(Global Load Balancing,GLB)等算法。In view of the unreasonable resource allocation and low resource utilization in the routing process of data center network,a path level-based genetic algorithm(PLGA)is proposed to improve the network transmission performance.With software define network(SDN),the Yen algorithm is used to generate the first generation population as the initial population.The initial population is initially optimized to choose the top K shortest paths.Three metrics are used to create a new fitness function,namely the available bandwidth quality of the path,the uniformity of the link-available bandwidth of the path,and the number of routing hops.The initial population is then subjected to selection,crossover,and mutation operations in order to identify the best path.Experimental results show that the PLGA algorithm outperforms equal-cost multi-path(ECMP),Multipath transmission-based dynamic load-balanced routing(MTDLR)in link bandwidth utilization,throughput and delay jitter performances.

关 键 词：软件定义网络 基于路径层面的遗传算法 动态路由 适应度函数 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]