基于相空间重构的Internet业务流量非线性动力学分析

基于相空间重构的概念,即通过测量函数获得的L维时间序列来研究动力系统的动力学行为的方法,研究了TCP-Reno在RED和DropTail机制下的拥塞窗口变化特性,以及TCP和UDP共享链路时的流量特性.利用相空间重构将单一标量时间序列重构为高维相空间,从而进一步研究了高维相空间的混沌等特性.分析结果表明:在一定的实验条件下,拥塞窗口变化值的关联维数为分数,最大Lyapunov指数为正,高维相空间具有混沌特性且其Hurst参数接近于1,此时的空间同时具有严格的二阶自相似性;流量值的关联维数为分数,最大Lyapunov指数为正,该空间亦具有混沌特性.

Analysis on Nonlinear Dynamics of Internet Service Traffic Based on Phase Space Reconstruction

The variability of TCP-Reno congestion window (CWND) under the RED and DropTail mechanism and the traffic trait of TCP and UDP with sharing links were studied based on the concept of phase space reconstruction which is a method studying behaviors of dynamic system through L dimensions time series that gained from the measurement function. A single scalar time series was reconstructed into a high-dimension phase space by the phase space reconstruction, then the chaotic characteristics of the high-dimension phase space was further researched. The results show that the correlation dimensions of CWND variable value under certain experimental conditions are fractional, the largest Lyapunov index is positive, the high dimension space has chaotic characteristics and its Hurst parameter is approximated to 1. Meanwhile, the space has strict second order self-similarity. The correlation dimension of traffic value is fractional, the largest Lyapunov index is positive, and the space also has chaotic characteristics.