Performance analysis of TCP in an IP network using fluid-flow models

There is a strong demand for Quality of Service (QoS) in the internet. One key element of QoS is traffic management during congestion. Congestion occurs when aggregate load exceeds the capacity on bottleneck link. In this paper, congestion is caused by heterogeneous traffic. A queue builds up in the router servicing the bottleneck link. We advocate RED (random early detection) as AQM (active queue management) policy to decrease the generated load when congestion detected and increase the load linearly in order to probe for available capacity when congestion subsides. The decrease and retransmission of packets due to drop persists decrease in TCP (transmission control protocol) performance. Large portion of the present work is to analyse and understanding of nature of CBR (constant bit rate) flows which are non-responsive to network traffic, on round trip time, TCP flows, packet losses, queue length. It is observed that CBR flows misuse cooperative nature of TCP flows. The unresponsiveness of CBR flows capture the bandwidth and make TCP starve without resources. The CBR flows benefit from their noncooperative behaviour and TCP flows suffer poor performance.