Research On Performance Evaluation And Improvement Of Spatial Information Network

In the development of spatial information networks,different access technologies,such as cellular networks,WiFi,D2 D,MESH,satellite networks and the Internet,will be integrated into a new heterogeneous network architecture to achieve global seamless connectivity.As a comprehensive network system,there are various types of application terminals and supported services of spatial information networks.It serves an eclectic object but is aimed at multiple users.The spatial information network effectively enhances the degree of integration and diversity of information.Aiming at the structural characteristics of spatial information network,this paper proposes to use segmentation modeling method and deep learning method to evaluate the network transmission delay.The segmentation modeling method uses the probability distribution function to simulate the delay of each sub-network,and the composite obtains the transmission delay of the spatial information network.The proposed deep learning method adopts the Convolutional Neural Network(CNN)model to learn the correspondence between network state,network structure,user service requirements and spatial information network delay.The CNN model can accurately evaluate and predict network transmission delay.Based on the large amount of index data obtained by OPNET simulation and its simulation results,the segmentation modeling method and the CNN model method have high accuracy and reliability,and each has its own advantages.This paper chooses the transport layer protocol MPTCP as the starting point for optimizing network performance.Aiming at the long round-trip delay and high packet loss rate of spatial information networks and the asymmetric performance between them,the improved strategy MPTCP-X is proposed in this paper.MPTCP-X draws on and combines the advantages of HayACK and TCP-Jersey to solve these problems.The improvement of MPTCP-X has three main aspects.The path of the slow path backhaul ACK is changed,so that the time interval at which the slow path receives the ACK packet is reduced,which helps to accelerate the growth of the transmission window.Calculating the optimal transmission window and accurately estimating the available bandwidth,so that the MPTCP sender can reduce the transmission rate in advance when the packet transmission rate is about to cause path congestion,effectively avoiding network congestion.The Congestion Warning(CW)mechanism is used to determine the specific cause of packet loss on the network,and different transmission window adjustment strategies are adopted according to network congestion and radio link errors.The NS2 simulation results show that MPTCP-X in heterogeneous satellite network state improves the end-to-end throughput of the network by more than 50% compared to TCP-Jersey,MPTCP and HayACK.The network performance evaluation method and protocol improvement strategy proposed in this paper have certain reference value for spatial information network research.