基子自适应泸波器理论的自相似流量预测  

Self-similar Traffic Prediction Using Adaptive Filtering

作　　者：纪其进[1] 

机构地区：[1]东南大学计算机科学与工程系,南京210096

出　　处：《通讯和计算机（中英文版）》2005年第4期42-49,共8页Journal of Communication and Computer

基　　金：高等学校博士学科点专项基金（No.20040286001）;国家973重大基础研究计划（No.2003CB31480）

摘　　要：根据流量特徵预测其到达速率是基于测量的网络控制机制的关键问题。本文研究了基于最小均方（Least Mean-Square，LMS）自适应滤波器对自相似流量进行速率预测的方法。通过分析对不同实际流量记录和仿真流量的预测结果，发现该方法不但减小了采用指数加权平均估计带来的计算复杂度，而且其滤波器系数自适应特性可以有效地跟踪流量的高度变化，从而更加准确地估计流量速率。此外，实验还得出了用LMS自适应滤波器进行流量预测的几个基本参数。One of the key issues in measurement-based network control is to predict the traffic arriving rate in the next control time interval based on the online measurement of traffic characteristics. Self-similar traffic prediction using least mean-square （LMS） adaptive filters is investigated in this paper. According to the prediction results of real and simulated traffic traces, it is found that the proposed method reduces the complexity significantly comparing with exponential-weighted averaging per packet, while enjoying more accuracy due to the adaptability of predicting coefficients. In addition, some key parameters of LMS adaptive filtering for self.similar traffic prediction are identified by experiments.

关 键 词：流量预测 自相似 自适应滤波器 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]