摘要
在研究自相似网络流量Hurst参数与相关长度之间的定量关系基础上,提出了基于小波变换的自回归预测的网络端到端流量预测方法,并利用预测值改进TCP拥塞控制算法。理论分析和仿真结果表明,基于小波变换的预测方法在网络流量预测方面比传统的时域方法有更高的精度。改进的拥塞控制算法在无线/有线混合网络环境中,相对于TCPW和Reno具有明显的性能改善。
After studying the relation between time series Hurst of network traffic and correlation length, an AR traffic prediction method in Wavelet domain was proposed, and then a novel network congestion control scheme was brought forward based on traffic prediction. Analysis and simulation show that Waveletdomain AR model gets good performance in high burstness traffic prediction and the performance of this congestion control scheme is better than TCPW and Reno in wired/wireless network.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第21期6935-6939,共5页
Journal of System Simulation
基金
国家863计划(重点)项目(2008AA062200)
关键词
拥基控制
流量预测
自相似
小波变换
IP网络
congestion control
traffic prediction
self-similarity
Wavelet
IP networks
