Research On Convergd Networking And Multipath Of Access Network

The evolution and extensive deployment of access network technology has promoted the vigorous development of services in Internet and mobile Internet.The emerging of novel network a architectures and new services such as the Internet of Things,Internet of Vehicles,Intelligent Manufacturing,and Virtual Reality lead to higher requirements in access networks.Such requirements are not only to increase bandwidth but also to reduce latency.Fiber and Wireless access network(FiWi)has a huge bandwidth in fiber sub-net and a flexible access style in wireless sub-net.Therefore,FiWi is an effective solution to meet the requirements of aforementioned novel services.In the early stage of development,there are two directions on converged networking research for FiWi access network.The first one is to study the networking architecture,resource scheduling and routing mechanism of the FiWi access network with the perspective of physical convergence.The other one is to solve the converged heterogeneous FiWi access networks problem by using network virtualization.Recently,some researchers have combined the access network and edge computing to study the converged networking strategies and discuss service mechanism to improve the network's ability for supporting low-latency services.However,the research on FiWi access network and edge computing only considers using the wireless sub-net as the access approach to offload the service,while using the fiber sub-net connects the end users and the central cloud.However,they do not integrate the fiber subnet and wireless subnet to unifiedly manage the resources.On the other hand,under the limited network resources,new converged access network is desired to have computing power to meet the demand of service,which are computing intensive and delay sensitive,and to ensure the transmission quality of the service,which requires the new converged access network with computing power.This paper is mainly supported by the National Natural Science Foundation of China.We apply the principle of network virtualization and combine edge computing to address the above issues.This thesis focus on access network converged networking and multi-path network virtualization.The major research contributions and innovations are as follows:1)The thesis proposes a converged FiWi access network architecture with edge computing.This thesis also present a resource scheduling algorithm with maximum revenue and a resource scheduling algorithm for central processing unit(CPU)balanced are proposed.Firstly,the proposed architecture includes a physical infrastructure plane and a control and management plane.The physical infrastructure consists of a FiWi access network and edge computing servers.All infrastructure resources are treated as virtual resources,and the control and management plane is implemented by SDN and NFV.In particular,the network,computing and storage resources are scheduled by NFV orchestrator and SDN controller.Secondly,the revenue-based resource scheduling algorithm(R-VNE)and CPU balanced resource scheduling algorithm(BCRA-VNE)are proposed.Specifically,the R-VNE sets as different price for for fiber links and wireless links.Given a determined processing requirements of the service,the fiber links are preferred by service.On the other hand,the BCRA-VNE proposes a metric in terms of computing resources balanced in the nodes.Based on such metric,this algorithm balances the computing resources in the network whenever a large number of services arrive.Lastly,we build an experimental simulation environment,and verify that the under the proposed architecture,the service acceptance ratio is increased by using the proposed architecture.Moreover,the simulation results show that the proposed algorithm has higher network utilization and higher revenue as compared with the existing resource scheduling algorithm.2)This thesis proposes a multi-path network virtualization scheme and a delay minimum multi-path network algorithm.Firstly,based on SDN and NFV technology,a multi-path network virtualization scheme is proposed.In the scheme,the network control and the data forwarding are separated to form a control plane and a data plane.The data plane includes a flow splitting,mapping,forwarding,and flow recovery modules to finish the multipath implementation steps,and the control plane is designed to monitor the network,network resource management,tag table maintenance,path selection and weight calculation modules,which provides necessary control management for multi-path design.Secondly,we conduct a test-bed based on the above scheme,and select five existing multi-path algorithms depolyed on the test-bed.The algorithms are compared from network utilization,throughput and latency to verify the effectiveness of the platform and solution.Lastly,based on network virtualization,a delay minimum multi-path network algorithm for elephant flow is proposed,which is compared with the Equal Cost Multi-Path(ECMP)algorithm.From the results,we find the proposed algorithm can reduce network latency and improve network throughput.3)This thesis proposes an implementation scheme of converged network for VR services and a multi-edge computing servers' resource allocation algorithm based on bankruptcy game.Firstly,the scheme is characterized in that the edge computing servers are physical proximity;the VR rendering process is placed in the edge computing servers;and the edge of the network is implemented by a high bandwidth and low latency of the converged access network.The edge computing servers cooperate to process the VR service.Secondly,the mathematical model of the VR service is presented,and the processing mechanism of the VR service under the cooperation of multiple edge computing servers is analyzed.The multi-edge computing server resource allocation algorithm based on the bankruptcy game is proposed to process the VR service.Lastly,we build an experimental simulation environment to verify the advantages of the proposed algorithm through parameters such as network throughput and VR service acceptance ratio.