Multi-dimensional Resource Allocation Algorithm For Radio Access Network Slicing

Various business application scenarios will be arised in the 5th-Generation Mobile Communication(5G),their key performance indicators are quite different while the traditional network architecture is highly coupled and cannot be flexibly customized.All of these restrict the development of 5G.Network Slicing(NS)technology based on Software Defined Networking(SDN)/Network Function Virtualization(NFV)is emerging and has become a key technology of 5G.Using network slicing technology,multiple logically isolated end-to-end networks can be virtualized on a physical network to meet different application requirements.As an important part of end-to-end network slicing,there is no clearly defined of RAN slicing architecture by the 3^rd Generation Partnership Project(3GPP),because of the complexity of it.This paper first studies the architecture of RAN slicing,then studies the multi-dimensional resource allocation algorithm,and finally verifies it through experiments.The main research includes:Drawing lessons from the 5G core network Service Based architecture(SBA),the RAN slicing architecture is designed to realize decoupling of network functions and standardized communication interfaces among network functions,And a RAN slicing experiment verification platform is builted based on the open source platform OAI(Open Air Interface),Flex RAN and Kubernetes.Secondly,the hard isolation and soft isolation between RAN slices and the bandwidth resource isolation scheme are designed,and the instantiation process of RAN slices is designed,too.Finally,the isolation and instantiation process of RAN slicing was tested on the RAN slicing experimental platform.The test results will show the feasibility of the RAN slicing design scheme and highlight the characteristics and advantages of network slicing.Aiming at the application scenarios that need to dynamically expand and shrink the slice resources during the specific RAN slicing service process,a joint optimization algorithm for computing,storage and network resources allocation is further studied.And the performance of the algorithm is verified based on the RAN slicing experiment platform built in this paper.The flexible scheduling of virtual computing and storage resources is realized by Kubernetes,and modify the virtual network resources of the base station in real time in a programmable manner is realized by Flex RAN.The algorithm combines simulated annealing algorithm and particle swarm algorithm to optimize the throughput of RAN slices and improve resource utilization under the premise of ensuring user Quality of Service(Qo S).RAN slicing experiment platform shows that the algorithm can dynamically adjust the multi-dimensional resources required by RAN slicing in real time and obtain an optimal resource combination to meet the Qo S requirement of users in RAN slicing.