Analyzing TCP Performance in High Bit Error Rate Using Simulation and Modeling

While Transmission Control Protocol (TCP) works well with a low bit error rate (BER), the performance of TCP degrades significantly if the BER rises above a certain level. A study of the performance of TCP with high BER is required for the efficient design and deployment of such systems. In this paper, we address the problem of TCP performance in high BERs and analyze the issues by investigating the effect of BERs on system performance. We consider TCP Reno in our study to explore the system performance using extensive analysis of simulation and modeling. In the analysis, we consider the amount of datagram sent and retransmitted, mean throughput, link-layer overhead, TCP window size, FTP download response time, packet dropping and retransmission, and the TCP congestion avoidance mechanism. We validate simulation results by setting up a virtualized testbed using Linux hosts and a Linux router. The results obtained show that TCP throughput degrades significantly and eventually collapses at the packet drop probability of 10% (BER = 10(-5)). The FTP download response time is about 32 times longer than that of a perfect channel (no packet dropping). We found that TCP Reno cannot handle such a high BER to operate in wireless environments effectively. Finally, we provide recommendations for network researchers and engineers confronted with the challenge of operating TCP over noisy channels.

Automated summary

This paper addresses the performance issues of TCP under high bit error rates (BERs) and analyzes the impact of BERs on system performance. Through extensive simulation and modeling, the study focuses on TCP Reno and explores various factors such as datagram transmission, throughput, link-layer overhead, window size, FTP download response time, packet dropping and retransmission, and congestion avoidance. The results demonstrate that TCP throughput significantly degrades and eventually collapses at a packet drop probability of 10% (BER = 10(-5)). Additionally, the FTP download response time is found to be 32 times longer compared to a perfect channel with no packet dropping. Recommendations are provided for researchers and engineers dealing with TCP operation over noisy channels.