Research On Flow Assignment And Data Scheduling Algorithms For Multipath TCP

The evolution of mobile communication technologies promotes the development of smart terminal and wireless service.Using the concurrent transmission ability of multiple network interfaces in a multi-mode terminal to provide users with more bandwidth and better service experience has being become a common topic among academic and industrial communities.Owing to the feature of the good compatibility with Transmission Control Protocol(TCP),Multipath Transmission Control Protocol(MPTCP)is a parallal transmission protocol and has attracted wide attention in recent years.However,due to the limited resource of terminal equipments,the utilization of the network resource reduces in heterogeneous wireless network,which makes the performance of the MPTCP degrade.In this paper,we investigate the flow assignment and data scheduling algorithms of the MPTCP with the aim to improve the transmission performance.For the problem of the receive buffer blocking resulting from the subflow dissimilarity in heterogeneous wireless network,we study the effect of the subflow dissimilarity,the packet loss rate and limited receive buffer capability on the performance of the MPTCP,in terms of the goodput.Moreover,we model the flow assignment of the MPTCP as an optimization problem,and propose an algorithm named MPTCP Goodput-Aware Flow Assignment and Data Scheduling(MPTCP-GA)to resolve the formulated problem.In the proposed MPTCP-GA algorithm,the flow assignment algorithm maximizes the MPTCP's goodput using a flow allocation mechanism,and the data scheduling algorithm distributes packets according to the result of the flow assignment and makes data packets reach the receive buffer in order.Simulation results show that the proposed MPTCP-GA algorithm can effectively deal with the deterioration of the transmission performance due to the subflow dissimilarity.It achieves larger goodput and lower average packet transmission delay.For the problem of the packet loss resulting from the unreliable wireless channel,we analyze the influence of network coding,the subflow dissimilarity and limited receive buffer on the goodput of the MPTCP with Network Coding(MPTCP/NC).The flow assignment with redundancy determination of the MPTCP/NC is formulated as an optimization problem.We propose the MPTCP/NC Goodput-Aware Flow Assignment with Code Redundancy Determination and Data Scheduling(MPTCP/NC-GA)algorithm to solve the optimization problem.In the proposed MPTCP/NC-GA algorithm,the flow assignment with code redundancy determination algorithm is aimed to maximize the goodput of the MPTCP/NC,and the data scheduling algorithm is utilized to distribute data packets reasonably.Simulation results show that the proposed MPTCP/NC-GA algorithm can further improve the transmission performance of the MPTCP in the heterogeneous wireless network.The proposed MPTCP-GA and MPTCP/NC-GA algorithms in the thesis can be applied to the network protocol stack to enhance the transmission efficiency of the MPTCP in real network.