论文部分内容阅读
近年研究发现网络中的业务量呈自相似特征,这种自相似特征显著影响网络的流量控制与排队性能,本文在自相似网络流量可预测的基础上,利用线性回归分析理论进行流量预测,并应用控制理论中的预测PI控制器原理设计了动态矩阵PI控制主动队列管理(Dynam icMatrix PI Control-Active QueueManagement,简称DM-PIC-AQM)算法,以克服队列的剧烈振荡,保持队列稳定在期望的长度.仿真实验结果表明,DMPIC-AQM算法在网络流量剧烈变化和小期望队列长度的情形下,DMPIC-AQM算法明显优于RED与PI算法.
In recent years, it has been found that the traffic in the network presents a self-similar feature. This self-similar feature significantly affects the traffic control and queuing performance of the network. Based on the predictable self-similar network traffic, this paper uses the linear regression analysis theory to predict the traffic and The principle of predictive PI controller in application control theory is used to design the dynamic matrix PI control-Active Queue Management (DM-PIC-AQM) algorithm to overcome the severe oscillation of the queue and keep the queue stable at the expected Length.The simulation results show that the DMPIC-AQM algorithm outperforms the RED and PI algorithms in the case of drastic changes in network traffic and small expected queue length.