Simulation-based Research On Self-similar Phenomena

Since the seminal study of Leland, Taqqu, Willinger and Wilson [3] which setthe groundwork for considering self-similarity an important notion in theunderstanding of network traffic, self-similarity has been the focus of recentresearch on traffic engineering. This dissertation addresses the cause of the self-similar network traffic. Based on the simulation with "Network Simulation-Version2" and wavelet analysis of simulation results, we demonstrate that a large numberof ON/OFF sources with heavy tailed distributed random ON and/or OFF periodsare not necessary to explain self-similarity, while TCP flow-control mechanism canbe a deterministic factor of self-similarity.This dissertation also provides an overview of recent development in networktraffic modeling with the notion of multifractal. It is also mentioned that theexisting technology for multifractal modeling of network traffic can also be usedfor measuring uncovering aspects of Internet performance that are of currentinterest.