Modelling Internet Traffic through Geographic Time and Space

Global Internet traffic can be modelled by a set of equations describing the periodicity of the time-based distributions of congestion. Congestion can be generated, firstly, by peak traffic from different time zones or secondly, by inadequacies within the regional network. The model is a dynamic form of spatial interaction modelling and can investigate the time delay of pings and packet loss at routers. The book will develop this model showing that it has characteristics peculiar to the Internet and develops the concept of relative time dependent on time zone origin of traffic. The first section of the book tests the model predictions in geographic time and space using data generated from a global network compiled by the Stanford Linear Accelerator Centre (SLAC) from 1998 to 2006. Examples are used from sites in the US, Europe, Asia and Australia to show the problems that the Internet has experienced in its evolution from either the 'tyranny of distance', an 'island effect' or inadequate investment in capacity. These sites can be classified relative to a regression line standardised relative to the rotation of the Earth to give an indication of the ability of the regional network to handle peak traffic flow periods. There is also a limit to transferring information by the Internet and this is expressed as an inequality underpinned by the speed of light. This will be again tested in the US and Europe. The second half of the book uses the model as a benchmark to study the evolution of the Internet from 1998 to 2006. Despite a significant rise in demand, the capacity of the Internet has increased, particularly within 1000km of the monitoring sites, from large scale regional investment in infrastructure. This investment has seen a considerable improvement in performance for the Internet from 2000 to 2004. The concept of relative time is developed where time-origins of traffic are considered as a complex variable with real and imaginary components. This can account for the bi- and tri-model distributions observed in the Stanford data. The socio-economic consequences of the model are discussed relative to stock market transactions, video games, and an Internet virus and the policy implications reviewed.