The Research Of Data Scheduling Based On MPTCP Over Asymmetric Multi-Path

With the development of network technology,more and more terminals have multiple network interfaces,but the traditional Transmission Control Protocol(TCP)can only use one interface for transmission,resulting in a waste of network resources.Using multiple network interfaces to achieve parallel transmission has become a new research.Multipath TCP(MPTCP)is a new multipath transmission protocol proposed by Internet Engineering Task Force,which can not only be compatible with existing TCP applications,but also support multiple network interfaces for multi-path transmission.However,the default data scheduling algorithm in MPTCP does not consider the complex network parameters,which will produce a large number of disordered packets in the heterogeneous network and reduce the transmission performance.In order to improve the transmission efficiency of MPTCP in asymmetric network,this thesis proposes a data scheduling algorithm based on delay and congestion status and a data scheduling algorithm based on reinforcement learning.The main works of this thesis are as follows:1.For the problem that the transmission performance of MPTCP is degraded in complex network environment,this thesis proposes a data scheduling algorithm based on delay and congestion.The proposed algorithm adds the consideration of network congestion factor based on the minimum Round-Trip Time(RTT)data scheduling algorithm.It calculates the proportion of RTT and congestion among different paths,and choose the path with small RTT and good congestion status.The simulation result shows that the proposed algorithm can improve the transmission efficiency.And then this thesis applies MPTCP to the actual production,and develops the Virtual Reality(VR)based remote interactive transmission system which provides VR live broadcasting and VR interaction.2.Combining the artificial intelligence algorithm of reinforcement learning with MPTCP data scheduling,this thesis put forward a data scheduling algorithm based on Deep Q Network(DQN).The proposed algorithm rewards and punishes each transmission path based on network information,dynamically estimates the transmission capacity of each path,and distributes the data to the path with strong transmission capacity.The simulation result shows that the proposed algorithm improves the transmission throughput and increases 20 percent amount of data compared with the traditional RoundRobin algorithm and the minimum RTT algorithm.