论文部分内容阅读
针对自相似流量具有的长相关和突发性特征,结合离散小波变换(DWT)和时间序列分析中的FARIMA过程,提出构建一种混合的网络流量预测模型来对网络行为的发展趋势进行分析和预测,并以预测结果辅助AQM的REM显式拥塞标记,从而提高流量控制的动态性以实现对传统网络协议策略的优化.实验表明该方案可有效改善协议性能,具有良好的稳定性和低丢失率特征.
Aiming at the characteristics of long-term correlation and suddenness of self-similar traffic and combining FARIMA process in discrete wavelet transform (DWT) and time series analysis, a hybrid network traffic prediction model is proposed to analyze the trend of network behavior Prediction and assists AQM’s REM explicit congestion flag with the prediction result to improve the dynamic of traffic control so as to optimize the traditional network protocol.Experiments show that this scheme can effectively improve the protocol performance with good stability and low loss Rate characteristics.