Research On Performance Enhancement Of Multipath Transmission For Heterogeneous Networks

With the continuous development of current networks,traditional single-path transmission protocols have gradually become difficult to meet user needs.Therefore,multipath transmission protocols are gradually proposed and applied to meet user demands for bandwidth,transmission stability and mobility,among which Multipath TCP(MPTCP)is widely regarded for its compatibility with existing networks and efficient transmission performance.However,in complex heterogeneous network environments,different path conditions and network environments vary greatly,which often have a significant impact on transmission performance.At the same time,different network environments often have different requirements for transmission protocols.For example,users usually require higher robustness in wireless mobile scenarios,and require low latency and high utilization of network resources in data center networks(DCNs).That also brings great challenges to the design of transmission protocols.This thesis focuses on the performance enhancement issues of MPTCP in different network scenarios,and provides different schemes to enhance MPTCP transmission performance.In this thesis,we focus on different network environments and application scenarios.First,the asymmetric environment in heterogeneous networks has significant differences in delay,bandwidth,and other parameters of different paths.And asymmetric paths of different subflows can lead to a heavy out-of-order problem,resulting in Head-of-Line(HoL)blocking and load imbalancing,which significantly degrades the performance of MPTCP.Secondly,wireless mobile networks are characterized by a high packet loss rate and frequent dynamic changes in the links.Traditional congestion control algorithms cannot achieve full bandwidth utilization in wireless mobile networks,and the overall transmission performance is often degraded due to the frequent condition changes in mobile scenarios.Finally,due to the high path diversity in DCNs,MPTCP does not always achieve the best results in terms of multipath selection and data allocation.When an overlap occurs between multiple paths of MPTCP,the resource utilization efficiency of the overall network will be degraded.At the same time,the non-optimal data allocation among multiple paths will also lead to the degradation of the overall performance.To solve the above problems,this thesis mainly proposes the following schemes:(i)In this thesis,we propose an MPTCP flow control scheme based on receive window allocation,which manages subflows based on the predicted buffer occupancy to control the data traffic.The proposed scheme achieves subflow-level flow control,which adaptively controls the data traffic on each subflow to improve the performance of MPTCP.By controlling traffic on each subflow individually according to the allocation of receive window,MPTCP can achieve load balancing and ensure resource allocation among subflows.At the same time,it can reduce the out-of-order packets caused by asymmetric paths and reduce the impact of HoL blocking on overall throughput.Experiment results show that the proposed scheme can dynamically adjust the resource allocation on subflows according to the real-time path conditions,improve the overall throughput and reduce the out-of-order packets in asymmetric networks.(ii)This thesis proposes Coupled BBR congestion control algorithm and AR&P adaptive scheduling algorithm to improve the transmission performance in wireless mobile network environments.Coupled BBR is a coupled congestion control algorithm for MPTCP to achieve both high throughput and stable sending rate in lossy network scenarios with guaranteed fairness with TCP BBR flows and balanced congestion.To further improve the performance,we propose an Adaptively Redundant and Predictive scheduler(AR&P),which includes two scheduling methods to improve adaptability in highly dynamic network scenarios and keep in-order packet delivery in asymmetric networks.Based on Linux kernel implementation and experiments in both testbed and real network scenarios,we show that the proposed scheme not only provides higher throughput,but also improves robustness and reduces out-of-order packets in some harsh circumstances.(iii)In this thesis,we study the rate allocation and multipath selection problems of MPTCP,and propose FMPTCP,which employs a feedback-based congestion control algorithm(FCC)and a feedback-based multipath routing algorithm(FMP)to jointly achieve high bandwidth utilization and low RTT in DCNs.FCC utilizes explicit congestion notification(ECN)to give accurate feedback of path congestion degree,and adjusts congestion window with gradient descent method to achieve optimal resource utilization and load balancing under a fixed routing topology.FMP utilizes a hop-byhop feedback approach to notify in-network congestion and path delay information,then provides transparent multipath routing for MPTCP flows.We show through extensive simulations that FMPTCP enables effective network resource utilization,which both improves the overall throughput and reduces transmission latency for DCNs.