基于改进布谷鸟搜索的Web集群自适应负载均衡算法  

Adaptive load balancing algorithm in web cluster based on improved cuckoo search

作　　者：张娜[1] 董亮亮 金瑜婷 包晓安[1] 吴彪 ZHANG Na;DONG Liangliang;JIN Yuting;BAO Xiaoan;WU Biao(School of Information Science and Technology,Zhejiang Sci-Tech University,Hangzhou 310018,China;Department of East Asian Studies,Yamaguchi University,Yamaguchi 753-8514,Japan)

机构地区：[1]浙江理工大学信息学院,杭州310018 [2]日本山口大学东亚研究科,日本山口753-8514

出　　处：《浙江理工大学学报（自然科学版）》2020年第4期527-534,共8页Journal of Zhejiang Sci-Tech University(Natural Sciences)

基　　金：浙江省重点研发计划项目(2019C03G2060359);浙江省自然科学基金青年基金项目(Q20F050032);浙江省公益技术研究计划项目(GG20F010028)。

摘　　要：为了解决Web集群中的高并发访问和资源异构引发的负载不均衡问题,提出了一种基于改进布谷鸟搜索的负载均衡算法。该算法建立了自适应负载分配权重模型,使用基于目标函数的布谷鸟搜索算法寻找最优权重。首先,根据集群任务调度特性定义了负载分配权重,并建立了集群调度的目标函数。其次,将负载分配权重编码为布谷鸟种群个体,使用布谷鸟搜索算法寻找最优个体,结合目标函数进行评估。另外,通过混沌变异增加初始种群的均匀度和离散度,并通过在布谷鸟搜索中引入反向学习,加速了最优权重的输出,根据最优权重将任务调度至集群中的各节点。结果表明:负载均衡算法使得整个集群的服务响应时间降低60%左右,在高并发请求情况下可显著提高异构集群的服务容量和负载均衡度。To solve the load-imbalance problem in the web cluster caused by high concurrent access and resource heterogeneity,a load balancing algorithm based on improved cuckoo search was proposed.For this algorithm,an adaptive load distribution weight model was established,and cuckoo search algorithm based on objective function was employed to solve the optimal weights.Firstly,load distribution weights were defined according to the characteristics of cluster task scheduling,and the objective function of cluster scheduling was created.Secondly,load distribution weights were encoded as individuals of cuckoo population,cuckoo search algorithm was applied to figure out the optimal individual,and assessment was made in combination with the objective function.In addition,the uniformity and disperse degree of the initial population were increased by means of chaos mutation,the output of optimal weights was accelerated by introducing opposition-based learning in cuckoo search,and tasks were assigned to nodes in the cluster based on the optimal weights.Experimental results showed that the load balancing algorithm reduces the service response time of the entire cluster by 60%or so,which can significantly increase the service capacity and load balance degree of heterogeneous clusters in the case of high concurrent requests.

关 键 词：布谷鸟搜索 反向学习 负载均衡 混沌变异 WEB集群 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]