Research On Network Slicing Resource Mapping And Orchestrating Based On SDN

With the digital transformation of the global industry and the booming development of the Internet,various new services are emerging.Promising services such as augmented reality,high-definition video,autopilot,smart grid,and telemedicine all these bring different performance requirements to the network.In order to support diverse service scenarios on the network and flexibly match the differentiated performance of diverse services,operators will meet the end-to-end performance requirements for service differentiation by providing flexible and customizable network slicing.As one of the core technology of 5G,network slicing is based on network virtualization.It creates multiple isolated end-to-end virtual networks on the physical infrastructure through service organization.Software-defined networking(SDN)promises flexible control of computer networks by organizing switches in the network dataplane through a centralized controller.The SDN architecture is applied for end-to-end network slicing,so that network slicing can provide flexible network resource allocation to meet various service requirements.This thesis scrutinizes and studies the SDN-based network slicing architecture,virtual network embedding in network slice lifecycle management,and network slice deployment.Through efficient virtual network embedding mechanism and network slice deployment orchestration mechanism to achieve a reasonable and efficient use of network resources to meet the diversified needs of future network.The main innovations in this article include the following aspects:(1)The SDN-based end-to-end network slicing architecture is enhanced,including the specific slicing architecture of the radio access network and the core network.Through the independent design of the access network and the core network architecture,the core network can support multiple heterogeneous access methods and meet the 5G system support for multiple access methods.This new architecture meets the diversification of future services by providing on-demand network slicings for different services,separating logical resources between different slices,and sharing hardware common infrastructure.(2)Aiming at the problems that the existing virtual network embedding algorithm solves the large-scale virtual network embedding,the efficiency of the algorithm is poor,the coordination of the nodes and links in the resource allocation is poor,the resource utilization is low,and the communication volume between the sub-virtual networks is large,a virtual SDN network embedding algorithm based on path splitting(PS-VNE)is propose.The PS-VNE algorithm first uses the k-region partition optimization theory combined with the adjacency partitioning method of the graph to divide a large virtual network request into multiple sub-networks in advance and then performs embedding in turn.In the node embedding phase of the sub-virtual network embedding,the resource status and the underlying topology characteristics are comprehensively considered,and the node embedding and link embedding phases are organically combined to improve the traditional two-phase embedding algorithm.At the same time,the support of path separability during the link embedding phase improves the virtual network embedding success rate and resource utilization.Simulation results show that the proposed new algorithm can effectively diminish the cost of virtual network embedding and improve the cost-to-benefit ratio.(3)For the problem of network slicing deployment,an integer quadratic programming model is formulated.At the same time,considering the high complexity of the algorithm and the long execution time for the exact solution of the model,it is not suitable for large-scale networks.Particle swarm optimization based network slicing deployment orchestration algorithm(PSO-NSDO).The PSO-NSDO algorithm uses particle swarm optimization to quickly converge to the characteristics of the global optimal solution.It also designs a fitness evaluation function that meets performance requirements for different types of business scenarios.Then,the particle swarm pursuing the global optimal solution and the local optimal solution are used to get the approximate solution of the network segmentation deployment problem.Finally,the simulation results show that the PSO-NSDO algorithm can provide corresponding deployment strategies for different use cases,meet the future network multi-scenario application requirements,and minimize the operator's cost and maximize the operator's revenue.