Research On Cross-layer Optimization For TCP Performance In Wireless Communication Systems

The increasing popularity of wireless networks which require high data rate and QoS-guaranteed traffic indicates that wireless links will play an important role in future internetworks. Reliable transport protocols such as TCP are tuned to perform well in traditional wired networks, where packet losses occur mostly because of congestion. However, networks with wireless links with characteristics of higher bit error rate, lower bandwidth, longer delay, frequent mobility and so on also suffer from significant losses due to bit errors and handoffs. TCP interprets these delays and losses as signs of network congestion and responds to all these losses by invoking congestion control and avoidance algorithms, resulting in degraded end-to-end performance in wireless systems. Currently, people have done a lot of research work to improve TCP performance in wireless environments including:(1) The majority of these schemes proposed are trying to optimize the system within the single layer. (2) To break the independence of the layers of OSI model, it aims to optimize the system by cross-layer design. Obviously, jointly considering the interaction between transport layer and other layers, it may bring the improvement for TCP.In this dissertation, under the guidance of information theory and optimization theory, the joint cross-layer optimization framework based on theoretical analysis and practical application for TCP protocol in wireless networks is proposed to ensure end-to-end QoS. The main contributions of this dissertation are as follows:Firstly, based on network utility maximization theory (NUM), QoS-guaranteed cross-layer optimization framework is proposed. The criteria for optimization of this framework, system model of joint optimization and optimization methods are introduced in detail.Secondly, multiple-hop wireless networks have a rapid development as a promising network technology in recent years. It will play an increasingly important role in wireless mobile networks in the future.multiple-hop wireless networks has the merits of high reliability and good scalability, but its multi-hop and dynamic self-organizing feature has brought enormous challenges to the application of multimedia services. In this paper, we proposed a systematic Network Utility Maximization (NUM) based cross-layer design approach using convex optimization methods for the joint optimization of physical, transport and MAC layers in order to maximize the network's overall throughput while at the same time providing end-to-end QoS for multimedia traffic. The proposed algorithm is fully distributed and scalable. Simulation results show that our cross-layer optimization can be performed iteratively based on local link information and obtain the network optimum TCP rate control and contention control parameters, which reduce the delay, maximize the system throughput while satisfying application's QoS requirements.Thirdly, in the standard TCP congestion control mechanism such as TCP Reno, it may cause the issue that packet sending rate increase exponentially in the slow start phase so significantly that it leads to the decline of TCP performance due to the overflow of bottleneck link buffer. One of the ways of TCP modification is based on estimation of the fair share of the available bandwidth for each flow and adjustment of the transmission window accordingly. This paper analyzes limitations of the standard TCP congestion control mechanism and proposes a novel TCP congestion control mechanism named TCP_ABE using adaptive bandwidth estimation in wired-wireless networks. Based on the estimation of the end-to-end available bandwidth, TCP_ABE dynamically adjusts the congestion window according to the congestion degree and the slow start threshold which builds a flexible slow-start parameter model and realizes the adaptive parameter setting. Simulation results show that the probability of multi-packet loss in the congestion window has been significantly decreased and TCP performance has been improved effectively by implementation of TCP_ABE.Finally, based on the analysis of TCP performance over IEEE 802.11 DCF, TCP flows becomes unstable when WLAN is in high load. In this paper, we propose a novel backoff algorithm for IEEE 802.11 DCF named as ACWB (adaptive contention window backoff) with cross-layer optimization. Its basic idea is to dynamically adjust the size of the contention window and packet length according to information interaction between MAC-layer and physical channel state of WLAN and alleviate the bad influence of MAC access mechanism on TCP congestion control. Simulation results show that ACWB algorithm can improve the TCP throughput and stability effectively.