Performance evaluation of enhanced-TCP in wireless IPv6 network

Abstract

The TCP congestion control algorithm is used to minimise the congestion in a network and to improve the performance of TCP. The TCP congestion control algorithm is originally designed to operate in wired network. When the TCP congestion control algorithm is implemented in a wireless network, the TCP congestion control algorithm assumptions are violated and result in performance degradation in a wireless network. This thesis presents the enhanced-TCP Vegas congestion control algorithm to optimise the performance of TCP in a wireless IPv6 network. Three enhancements are proposed in the enhanced-TCP Vegas algorithm. The first enhancement is to minimise the bursty traffic flow at the TCP sender network. The first enhancement is called Packet Spacing. The second enhancement is to optimise the probing of the transmission rate into the network. The second enhancement is called Progressive CWND. The third enhancement is to differentiate between the real congestion and the spurious congestion in the wireless network. The third enhancement is called Congestion Measure. The proposed algorithm is implemented in ns-2. The simulation experiment is conducted in two scenarios. The first scenario investigates the performance of TCP Vegas algorithms in single TCP connection network. The results of simulation experiment show that the enhanced-TCP Vegas algorithm, which combines the packet spacing, progressive CWND and congestion measure methods performs compared with the other TCP congestion control algorithms that are used in this research. The second scenario investigates the performance of the enhanced-TCP Vegas in multiple TCP connections. The results of simulation experiment show that: (i) as the number of TCP connection in a network increases, the packet delay, network throughput and network efficiency increases; (ii) the throughput fairness increases as the packet size increases. However, as the number of TCP connection increases, the variation of the throughput fairness for all the packet sizes decreases. The proposed algorithm shows poor throughput fairness against standard TCP Vegas algorithm; (iii) as the number of TCP connections in network increase, the handover recovery time increases accordingly. The proposed algorithm exhibits lesser handover recovery time compared to the standard TCP Vegas algorithm; (iv) the appropriate packet size of FTP file is inversely proportional to the number of TCP connection in a network; (v) as the buffer size in a network increases, packet delay, network throughput and network efficiency increases. Additionally, this research presents the appropriate packet size and buffer size to be used over the multiple TCP connections network by using the proposed algorithm.