Study On Communication And Computation Resource Allocation For 5G End-to-end Network Slicing

In the 5G(5th Generation)mobile network,to support business diversity and meet the differentiated needs of vertical industries,network slicing technology has emerged.To better meet various vertical industry applications,network slicing is required in both the core network and the radio access network,that is,to achieve end-to-end network slicing.Network slicing can customize network functions and allocate network resources on-demand independently according to the needs of business scenarios.In a network slicing-based network,operators need to efficiently map the network slicing request to the underlying physical network after receiving a network slicing request,which involves the optimal allocation of physical resources.Maximizing the use of multi-domain resources under the network slicing architecture,including communication,computing,and storage resources,can minimize deployment costs and capital costs,and optimize performance such as latency and energy consumption.This thesis is based on the cooperation project of China Unicom Network Technology Research Institute "B5G Network System Key Technology Research Project".The thesis addresses the joint allocation algorithm of communication and computation resource for 5G end-to-end network slicing.Considering static resource allocation scenario and dynamic resource allocation scenario,joint allocation algorithms for static and dynamic communication and computation resource are proposed respectively.The main work of this thesis is as follows:1)The researches of 5G end-to-end network slicing resource allocation is reviewed.First,this thesis gives an overview of 5G end-to-end network slicing,including end-to-end network slicing,core network slicing and radio access network slicing related concepts and main technologies.Then we summarize the classification of mobile network slicing resource management research.Finally,we review the current research status of mobile network slicing resource allocation,comb and summarize the current research status of core network slicing,radio access network slicing,and end-to-end network slicing resource allocation,point out the problems faced by current research and clarify the research direction of our thesis.2)In 5G end-to-end network slicing,aiming at the low latency requirement of static URLLC(Ultra Reliable and Low Latency Communication)network slicing,a communication and computation resource joint allocation algorithm based on subgraph isomorphism is proposed.This thesis proposes a framework of end-to-end mobile network slicing.We model an end-to-end delay minimization problem based on the proposed framework.On this basis,the optimization problem of network slicing end-to-end delay is established with VNF(Virtualization Network Function)mapping in the core network and radio resource allocation in the access network as optimization variables.The proposed algorithm gives the suboptimal solution to the proposed optimization problem through multiple iterations and pruning in the graph matching algorithm.The simulation results show that the proposed algorithm can effectively reduce the end-to-end delay of network slicing,and guarantee the reliability requirements of network slicing as well,and meet the requirements of low-latency and high-reliability services.3)In 5G end-to-end network slicing,a DQN(Deep Q Network)-based two-stage joint allocation algorithm for communication and computation resource is proposed to solve the problem of dynamic changes of network slicing data queues,radio channel status and physical network topology.The dynamic resource allocation model of end-to-end slicing is constructed,and the dynamic joint allocation of communication and computation resource is carried out to maximize the overall utility of the network on a long-term scale.The dynamic migration of VNF and the flexible allocation of virtual network resources are realized according to the service state and QoS(Quality of Service)requirements of virtual network users.In the proposed algorithm,the core network-side VNF migration strategy and mapping method obtained by the one-stage algorithm are taken as the input of the two-stage algorithm to obtain the access network-side CPU(Central Processing Unit)resources and PRB(Physical Resource Block)allocation result.The simulation results show that the proposed algorithm can optimize the overall utility of the network on a long-term scale,improve the long-term average revenue,and reduce the average cost of the system.