Effect of Traffic Self-Similarity on Network Performance

The results of rtaditional Markovian traffic model differ greatly with actual traffic in network performance. The reason is that there is self similarity in network communication. Bursts can happen in any range of thime. Its buffer requirement is far greater than that of Poisson distribution. This paper reveals features and underlying cause of self similarity traffic, investigates the effect of self similarity traffic on network performance, and presents an approach to predict traffic with probability to improve the network utilization rate.