基于需求预测的云无线接入网计算资源分配策略研究

Research on computing resource allocation strategy for cloud radio access network based on demand forecasting

云无线接入网利用网络功能虚拟化和软件定义网络技术以支持端到端的网络切片,使得接入网可共享无线、终端和网络等资源,已成为5G网络中优先采用的网络架构。针对云无线接入网中端到端的网络切片场景,通过控制平面建立的数据驱动运维框架来收集网络信息并进行数据处理。预测未来一段时间内计算业务量的需求,设计了一种基于虚拟化网络功能(VNF)的计算资源分配方案,提出了基于降序最佳适应(BFD)的离散粒子群算法。仿真结果表明,所提出的策略和算法可实现云无线接入网计算资源的动态灵活分配,并能有效降低VNF迁移能耗和迁移次数。

Cloud radio access network(CRAN)is a new network architecture commonly used in 5G network.It uses network function virtualization and software defined network technology to support end-to-end network slicing so that access network can share the same infrastructure.A computing resource allocation scheme based on virtualized network function(VNF)computing resource demand forecasting was proposed for cloud wireless network supporting end-to-end network slicing.Data-driven operation and maintenance framework established by the control plane collects network information and performs data processing to predict the demand for computing traffic in a period of time,which used the discrete particle swarm optimization based on best-fit-decreasing to dynamically and dynamically allocate CRAN computing resources,reducing the migration energy consumption and migration times of VNF.