All current versions of reliable Transmission Control Protocol (TCP) react to packet losses differently and adjust the TCP congestion window in various ways. These protocols assume congestion in the network to be the primary cause for packet losses and unusual delays. TCP performs well over wired networks by adapting to end-to-end delays and packet losses caused by congestion. The TCP sender uses the cumulative acknowledgements it receives to determine which packets have reached the receiver, and provides reliability by retransmitting lost packets. The sender identifies the loss of a packet either by the arrival of several duplicate cumulative acknowledgements (say, three ACKs) or the absence of an acknowledgement for the packet within a timeout. TCP reacts to packet losses by reducing its transmission (congestion) window size before retransmitting packets, initiating congestion window or avoidance mechanisms and backing off its retransmission timer. These measures result in a reduction in the load on the intermediate links, thereby controlling the congestion in the network. Unfortunately, when packets are lost in the networks for reasons other than congestion, these measures result in an unnecessary reduction in end-to-end throughput and sub-optimal performance.
Wireless links typically have much higher bit error rates. This implies that packet loss would occur frequently. If no error correction is attempted at lower layer, TCP will exercise its congestion control procedure unnecessarily and the throughput will be reduced significantly. If the link layer performs error control by performing the retransmission itself, packet transmission time will vary greatly, sometime even exceeding TCP retransmission time out and again TCP slow start will occur. In wireless networks, “packet loss ’’ problem is also encountered during handover when a mobile device moves from the coverage of one cell to that of another. During the handover, if the mobile station decides to make a handover before the segments are transmitted over the air interface, it is likely that some TCP segments buffered in a base station may be forwarded to another base station. This results in excessive segment delay or loss.
Thus, there is a clear demand for methods that can suppress the problems caused by the wireless environment. Recently, several techniques have been developed to improve end-to-end TCP performance over wireless links. They can be classified into three categories: end-to-end TCP, split TCP and link layer TCP. However, they have not addressed these problems successfully.
In this thesis, we propose, design and implement several algorithms that are applicable to the wireless networks in order to solve outstanding problems. Firstly, the research investigates the relationship between packet loss and network congestion and introduces a feedback based end-to-end congestion control algorithm to the wireless network. This algorithm is a modification of a Fair Intelligent Congestion Control (FICC) proposed in . The innovation of the algorithm is to modify the original FICC in such a way that the queue lengths can be effectively controlled when it is jointly employed with TCP in the wireless network.
The next algorithm is the new design of Explicit Loss Notification (ELN) at base station in Wired-Cum-Wireless networks. With the combination of new ELN algorithm and Wireless FICC algorithm, the end-to-end performance and fairness are greatly improved by eliminating the misinterpretation of error related lost packets from congestion.
Finally, the research investigates the effects of network congestion, which often happens over low bandwidth wireless link, and QoS performance (e.g. fairness, delay variation) of multiple sessions of TCP traffic in a hybrid network. We propose a framework, which consists of two main algorithms, feedback based congestion control and Explicit Window Adaptation (EWA).