Chaotic network attractor in packet traffic series

This paper extracts the chaotic network attractor from actual measured packet traffic series of TCP type. Both the analysis on aggregated traffic series and packet arrival interval series can show similar chaotic properties. Further investigation on different time scales shows that the attractor can exhibit different geometric shapes. Principal component analysis (PCA) on the phase space reconstructed shows that the dominant dimensions count decreases as time scale increases. When the time scale is small enough, there are at least 3 or 4 state variables dominate the network dynamic behavior while only one at large time scales. Same tests on UDP traffic series also show similar chaotic behaviors but with slightly difference. Comparison between actual measured traffic and surrogate traffic generated by four popular traffic models shows that the current widely used traffic models like the Poisson arrival traffic model, the deterministic chaotic-map model, the FBM-based model and the independent wavelet model cannot capture such network behaviors.

[1]  H. Abarbanel,et al.  Determining embedding dimension for phase-space reconstruction using a geometrical construction. , 1992, Physical review. A, Atomic, molecular, and optical physics.

[2]  Richard G. Baraniuk,et al.  Simulation of nonGaussian long-range-dependent traffic using wavelets , 1999, SIGMETRICS '99.

[3]  Chuanyi Ji,et al.  Modeling heterogeneous network traffic in wavelet domain , 2001, TNET.

[4]  H. Kantz,et al.  Nonlinear time series analysis , 1997 .

[5]  Walter Willinger,et al.  On the self-similar nature of Ethernet traffic , 1993, SIGCOMM '93.

[6]  L. Cao.  Practical method for determining the minimum embedding dimension of a scalar time series , 1997 .

[7]  Parag Pruthi,et al.  Chaotic Maps As Models of Packet Traffic , 1994 .

[8]  Walter Willinger,et al.  Self-similar traffic and network dynamics , 2002, Proc. IEEE.

[9]  H. Takayasu,et al.  Dynamic phase transition observed in the Internet traffic flow , 2000 .

[10]  Richard G. Baraniuk,et al.  Multiscale queuing analysis of long-range-dependent network traffic , 2000, Proceedings IEEE INFOCOM 2000. Conference on Computer Communications. Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies (Cat. No.00CH37064).

[11]  Andras Veres,et al.  The chaotic nature of TCP congestion control , 2000, Proceedings IEEE INFOCOM 2000. Conference on Computer Communications. Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies (Cat. No.00CH37064).

[12]  Walter Willinger,et al.  Self-Similar Network Traffic and Performance Evaluation , 2000 .