Research On Heterogeneous Multipath Collaborative Transmission Mechanisms For Smart Integration Identifier Networking

In recent years,network technology has developed rapidly.The multi-interface intelligent terminal device is becoming more and more popular.Integrating heterogeneous network resources for multi-path collaborative transmission has become a key technology to support high data volume and high reliability transmission services.However,the heterogeneous network links show strong diversity and lack of stability in nature.The static and rigid original design of traditional networks leads it to drawbacks in heterogeneous networks integration,intelligent resource adaptation,and diversified services guarantee,which is unable to provide strong foundation for heterogeneous multi-path collaborative transmission mechanisms.Therefore,research on new network architecture to support efficient and heterogeneous multi-path collaborative transmission mechanisms has become one of the hot topics in the world.Smart integration identifier networking(SINET-I)is a new network architecture.In this architecture,network resources are intelligently integrated from multi-space,multidimension and the whole network domain.The performance of heterogeneous multi-path collaborative transmission can be effectively guaranteed,which also provides an infrastructure for realizing efficient and reliable multi-path collaborative transmission.Based on the SINET-I architecture,this thesis carries out key technical research from three aspects: transmission process,sending process and receiving process.The research is focused on the goal of multi-path collaborative transmission,and emphatically on the problems of network model,congestion control,packet disordering,and receive-side buffer blocking,which are faced in heterogeneous multi-path collaborative transmission.The main innovations and contribution of this thesis are as follows:(1)Based on the analysis of the heterogeneous multi-path collaborative transmission model in the SINET-I architecture,the research clarifies the overall framework and network model of the heterogeneous multi-path transmission based on the SINET-I.Firstly,this thesis briefly describes the overall architecture of ”three-layer and three domain”,analyzes the cooperative relationship between heterogeneous multi-path collaborative transmission mechanisms in all domains,and proposes a general solution for multi-path collaborative transmission under the SINET-I.Subsequently,based on the test data the research clarifies in high-speed mobile scenario,a network model based on two time scales characteristic is proposed.On this basis,the advantages of window uncoupled congestion control algorithm in high-speed mobile scenarios are analyzed,which lays a theoretical foundation for subsequent research.(2)To solve the congestion control management problem in heterogeneous multipath collaborative transmission,a transmission component function model based on SINET-I is designed,and a multi-path collaborative congestion control mechanism based on window uncoupled is proposed.Firstly,the congestion window traffic problem in multi-path collaborative transmission is described,and an optimization algorithm based on the α-fair utility function is established.On this basis,this thesis proposes a window uncoupled congestion control algorithm based on reinforcement learning,which can realize fine-grained adjustment of congestion window size according to different scenarios.The simulation results show that the proposed algorithm can effectively improve the stability of parallel multi-path transmission in the high-speed rail communication network,and improve the overall throughput of multi-path cooperative transmission.(3)Aiming at the problem of packet disorder in heterogeneous multi-path collaborative transmission,this thesis designs a forwarding component function model based on SINET-I,and propose a multi-path cooperative data packet scheduling mechanism based on optimized order.Firstly,the packet scheduling problem existing at the sender in multi-path collaborative transmission is described,and a packet scheduling optimization algorithm based on path state awareness is established.Then,a programmable multipath packet scheduling framework is implemented,and the flexible packet forwarding is realized by using the programmable data plane;on this basis,a multi-path data packet scheduling algorithm based on reinforcement learning is proposed,and the asynchronous learning framework is used to improve the training efficiency of the algorithm.Simulation tests and system implementation show that the proposed algorithm can effectively alleviates packet disorder under the multi-path collaborative scheduling mechanism while improving the throughput.(4)Aiming at the problems of buffer blocking and reliability at the receiving end in heterogeneous multi-path collaborative transmission,a functional model of network coding/decoding components based on SINET-I is constructed.Firstly,the buffer blocking problem at the receiving end in the heterogeneous multi-path collaborative transmission is described,and the optimization problem based on the goal of minimizing the redundant coding rate is established;Then,a programmable multi-path adaptive network coding framework is implemented,which utilizes the programmable data plane to realize flexible coding of data packets;On this basis,a multi-path adaptive network coding algorithm based on reinforcement learning is proposed.The simulation test and the system implementation verify that the proposed algorithm improves the reliability of the multi-path collaborative transmission while alleviates the buffer cache problem effectively.