数据中心流量调度的分簇聚类算法仿真  

Simulation of Cluster Clustering Algorithm for Data Center Traffic Scheduling

作　　者：屈晓[1] 刘海[2] QU Xiao;LIU Hai(Zhujiang College,South China Agricultural University,Guangzhou Guangdong 510900,China;School of Computer Science,South China Normal University,Guangzhou Guangdong 510631,China)

机构地区：[1]华南农业大学珠江学院,广东广州510900 [2]华南师范大学计算机学院,广东广州510631

出　　处：《计算机仿真》2023年第6期513-517,共5页Computer Simulation

基　　金：2021年广东省高等教育教学研究和改革项目(粤教高函[2021]29号)。

摘　　要：为了避免数据中心出现信道拥塞问题,需要对数据中心中的流量展开调度处理,为此提出面向数据中心流量调度的分簇聚类算法。分析数据中心的拓扑结构,检测网络中存在的流量数据,通过梯度下降方法对多层感知机的参数展开调节,利用调节后的多层感知机对检测到的流量数据展开去噪处理,提升流量数据分簇精度。采用K-means算法分簇聚类处理去噪后的流量数据,通过网络带宽分配,在相关约束条件的基础上分簇调度流量数据,实现数据中心流量的调度。实验结果表明,所提方法的分簇精度较高,流量包的速率基本相同,表明所提方法具有较高的稳定性,调度效果较好。In order to avoid the channel congestion problem in the data center,it is necessary to schedule the traffic in the data center.For this reason,this paper puts forward a clustering algorithm for data center traffic schedu-ling.Firstly,the topology of data center was analyzed,and then the traffic data in the network was detected.Secondly,the parameters of the multilayer perceptron were adjusted by the gradient descent method.Based on the multilayer perceptron,the noise was removed from the detected data,so that the clustering accuracy of traffic data was improved.Moreover,the K-means clustering algorithm was used to process the de-noised data.On the basis of related con-straints,traffic data was clustered and scheduled through the network bandwidth allocation.Finally,the data center traffic scheduling was achieved.Experimental results show that the proposed method has high clustering accuracy and similar rate of traffic packets,indicating that the method has high stability and good scheduling effect.

关 键 词：数据中心 多层感知机 流量调度 网络带宽分配 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]