Research On Highly Robust MPTCP Adaptive Data Transmission Technology

As the Internet and related industries in the global have been growing rapidly, the amount of network application business has become much larger. However, traditional TCP protocol can only use an interface for single-path transmission, which is difficult to meet the requirements of high reliability and high efficiency in large traffic applications. Multi-path parallel data transmission is able to establish multiple parallel paths by multiple communication interfaces of the intelligent terminal, which is beneficial to improve transmission performance. MPTCP (Multipath TCP) protocol made improvements to TCP, to achieve a multi-path parallel data transmission technology, which has better robustness and transmission efficiency than traditional TCP. MPTCP needs to establish and maintain multiple heterogeneous network paths.Because of the large differences in performance of various types of network paths, if one or more poorly performing paths cause network fluctuations, the overall network fluctuation may occur, resulting in packet loss, out-of-order phenomenon, network congestion and other anomalies. At present,there are few relative researches on MPTCP data transmission strategies and algorithms under heterogeneous network and network fluctuation. Therefore, this thesis studies the adaptive data transmission technology of MPTCP, which can adaptively adapt to heterogeneous network and network fluctuation scenarios to adjust the transmission strategy, so as to achieve high reliability and high transmission throughput.Firstly, a congestion control algorithm based on Markov prediction model was designed in this thesis,which optimized the existing congestion control of MPTCP,so that the optimized MPTCP is able to make smooth and effective feedback and change for network congestion,packet disorder and packet loss in the process of transmission.Simulation results show that the proposed congestion control algorithm increased the overall average throughput by about 10% compared to other existing MPTCP congestion control algorithms, and the performance curve is smoother, which means better stability. The Advantage is more obvious especially in the case of large packet loss rate,severe jitter and small transmission buffer at the receiving end. Secondly,a dynamic network coding strategy based on Raptor code is designed. The systematic Raptor code is introduced into MPTCP data transmission, which makes MPTCP data transmission obtain the capability of error correction and data restoration while maintaining high performance. Simulation results show that the dynamic network coding strategy proposed in this thesis has a smaller throughput variation and a 15%-improvement in overall throughput compared to standard MPTCP. Compared with other MPTCP network coding strategies, the overall throughput is improved by 5%, or improved by 10% in the case of poor network quality, moreover transmission jitter is smaller, which means data transmission performance is more stable.The results of this thesis might provide theoretical contributions for MPTCP research on heterogeneous networks and network fluctuation scenarios, and is able to provide reference and guidance for engineering practice of MPTCP congestion control and network coding.