一种基于支持向量回归的互联网端到端延迟预测算法

互联网端到端延迟是指IP分组沿着互联网中一条确定路径进行传输的延迟,端到端延迟的精确预测是大量网络活动的基础,从网络协议设计到网络监测,再从确保端到端QoS性能到各种实时业务性能提升。提出一种新的端到端延迟的预测方法,主要贡献有:a)将互联网端到端延迟预测的问题转换为多元回归的预测问题,提出了基于多元回归的端到端延迟预测框架;b)采用支持向量回归SVR方法来求解端到端延迟的多元回归问题,提出了基于SVR的互联网端到端延迟预测算法。最后使用互联网采集的RTT数据来验证提出的算法,实验结果表明,提出的预测算法具有快速和精确特点,是一种适合实际应用的预测算法。

Internet end-to-end delay prediction using support vector regression

End-to-end packet delay of the Internet is the IP packet transmission delay along a determined path. An accurate end-to-end delay prediction is fundamental to numerous network activities, from protocol design to network monitoring, and from ensure end-to-end QoS to performance enhancement for realtime network applications. This paper presented a novel methodology for predicting end-to-end delay. The major contributions are: a) It converted the end-to-end delay prediction problem into the multivariate regression, and proposed a multivariate regression-based forecasting framework for end-to-end delay; b) It employed support vector regression (SVR) to solve the multivariate regression problem of end-to-end delay, and induced a SVR-based end-to-end delay predicting algorithm. Finally, it used the actual RTT data collected from Internet to validate the proposed algorithm. Simulation results show that the proposed algorithm has fast and accurate prediction characteristics, which is very suit for practical applications.