Abstract:
The volume of Internet Traffic has experienced an explosive growth over recent decades. Over the last twenty years TCP has become the dominant application layer protocols and most Internet traffic is TCP traffic. More and more applications are now migrating towards over TCP other than UDP. Therefore, it is important to understand the long term trends and behaviors of TCP flows in order to better understand TCP traffic. One important objective of this project is to design, development and analyze TCP traffic traces over different years and different organizations to investigate the long term impacts of TCP flows in network traffic trace files by using four traditional measurement metrics: the average packet length of TCP flows, the lifetime of TCP flows, the packet count of TCP flows and the Round-Trip Time of TCP packets. The second objective of this thesis is to observe different features of individual TCP flows by passively extracting samples of long lifetime TCP flows from each traffic trace files. We have observed that some TCP flows appear to have a regular pattern of RTT and RTT Variation distribution. We have also noted that some long active TCP flows with small volume transmitted have relatively fast response times. Self-Similarity appears to exist in the RTT and RTT Variation distribution of individual TCP flows, and TCP flows with long RTT might be result of various factors.