A decade of Internet research - advances in models and practices

The complexity and heterogeneity of the current Internet have rendered traditional analytical models and techniques inadequate for networking researchers and engineers. Many within the networking research community feel that researchers investigating new protocols and architectures, either by simulation or test-bed implementation, need to use common models. Despite the lack of models with universal applicability, there are certain models that are more appropriate than others for analysing certain systems. Researchers in different areas tend to use established models, typically to allow comparison of results. In addition, researchers have made considerable progress in understanding the statistical nature of Internet traffic. Despite the widespread use of simulation and test-bed implementation for modelling the Internet, there is little consensus today on analysis techniques and their validation. This paper summarises the main advances made in the last ten years or so in understanding the nature of Internet traffic, models and practices developed for Internet topology and protocol dynamics analysis.