The Impact of Self-Similarity on Network Performance Analysis

Recently, statistical analysis of high-resolution measurements of several types of network tra c has shown that many type of network tra c are self-similar or fractal in nature. This report gives an overview of self-similarity and examines the e ect of self-similar inputs on the performance of asynchronous transfer mode (ATM) switches. This is done through the use of trace based simulation using actual Ethernet tra c traces. The results of these simulations suggest that self-similarity has an adverse e ect on the performance of ATM switches and that certain suggested bandwidth allocation policies for ATM switches will signi cantly outperform others in the presence of self-similarity. ii Acknowledgements I would like to acknowledge several people who have helped with this project: Mike Devetsikiotis for providing me with numerous papers addressing the topic of selfsimilarity and for introducing me to several members of the Bellcore research team, Rob Tyson for providing me with access to the Splus sofware used for generating synthetic trace data, Vern Paxson for several email discussions regarding the FFT method of generating self-similar sample paths, and Kevin Fettig for giving me pointers to some papers on self-similarity. Finally, I would like to thank my supervisor Dr. John Neilson for numerous helpful discussions as well as for having introduced me to the eld of computer systems performance analysis. iii