一种基于噪声模型的TCP有效带宽估计方法研究(英文)

Method of Available Bandwidth Estimation for TCP Based on Channel Noise Model

准确的带宽估计和网络建模对于网络拥塞控制算法的设计具有重要的意义。采用通信系统加性噪声分析方法和信道容量理论,将TCP链路背景流量的影响等效为加性噪声,提出建立了一种基于等效噪声的链路模型和带宽估计算法。该模型不同于传统带宽估计对采样值直接进行滤波处理来得到估计带宽,而是利用网络流量的随机过程特性采用统计方法进行有效带宽估算,所得到的估计带宽有界并且具有良好的收敛性和准确性。理论分析和仿真结果表明了该模型和算法是正确的。

The accurate bandwidth estimation and network model make significant sense for the design of network congestion control scheme. Introducing methodology of noise analysis and channel capacity theory in communication system, a TCP path was regarded as an additive noise channel, where background traffic was considered as noise, then link model and bandwidth estimation algorithm were proposed based on the equivalent noise. The remarkable difference from traditional bandwidth estimation methods is that the algorithm does not utilize the direct filtering the samples, but makes use of stochastic characteristics of network traffic and statistical analysis. At the same time, the bandwidth estimated is bounded with better accuracy and convergence. Theoretical analysis and simulation show that the link model and estimation scheme can work well.