Chaos and Predictability of Internet Transmission Times
Joseph C. Park
Electronic mail address: josephpark@bellsouth.net.
Zenith, Inc.,
3045 Church Hill Drive,
Boynton Beach, FL, USA
Abstract
The Internet consists of millions of interconnected network nodes comprising a complex system utilized for information storage, processing, and transmission. It is demonstrated that the temporal evolution of transmission times for an information packet transmitted between two stable endpoints across the global network constitutes the chaotic time series of a dynamical system with positive Lyapunov exponents and fractal dimension. Examination of system invariants establishes the predictability of local dynamical variables and sets bounds on the ability to forecast the temporal evolution of the system variables. An artificial neural network (ANN) is invoked to learn and predict the Internet response times, wherein it is established that even though the system is dynamically divergent from a local system variable perspective, the ANN is fully capable of characterizing and predicting the macroscopic behavior of the global Internet network response times.
