Comparison of Web and KAZAA Traffic on Self-similarity: A Case Study

J. Liu and J. Copeland (USA)

Keywords

Selfsimilarity, network traffic, web and KAZAA.

Abstract

The essential characteristic of network traffic is self similarity (SS). Understanding the self-similarity of data traffic is important to network engineering. In the earlier studies, data traffic is described as an on-off process and the heavy-tailed file sizes have been used to project long range dependence (LRD) the traffic self-similarity at large timescales. In this work, we investigated two types of traffic from web and KAZAA applications. We observed that the aggregated TCP traffic is consistent with the former results exhibiting SS and LRD in the full range, and the web traffic shows a well-defined scaling region at large timescales. However, KAZAA traffic has a limited scaling range only at the small timescales (below 100 ms), despite KAZAA is more heavy-tailed in the connection size than web traffic. This discovery motivates us to examine the differences in the application behaviors on the traffic self-similarity. Furthermore, we have developed a novel method to measure traffic bursts on self-similarity in the data domain. We used the traffic burst method to profile the Web and KAZAA traffic traces and demonstrated that this method could be used to characterize the self-similar traffic flows in the network traffic engineering.

Important Links:



Go Back