Abstract: Here presented is a great discovery in the area of performance evaluation of high speed network since 1990,i.e., the real traffic has self similarity. The consequence of the discovery is that the conventional Markov models are not applicable to network traffic modeling. New models and tools have to be developed. First given is a mathematical definition of self similarity,and then followed is a discussion of the properties of self similarity. The relationship of heavy tail distribution and self similarity is also discussed. Also introduced are several methods of generating self similar traffic data and two famous results of queueing analysis based on self similar traffic.Finally, some of the work under development is mentioned.