面向白盒交换机的软件定义网络切片测试平台及验证

Software-defined network slicing test bench and validation for white-box

摘要:

随着电力行业的高速发展,人们对电力有着更多的需求,一些重要的科研项目也在朝电力方向不断深入,但当前的电力物联网缺乏由白盒交换机搭建的SDN切片测试平台。因此,本文提出了面向白盒交换机的软件定义网络切片测试平台,首先,设计了一种融合数字孪生(DT)、AI智慧内生的SDN切片平台架构；其次,通过对白盒交换机、链路数据的统一表征,构建了孪生网元本体模型、链路本体模型和与物理网络一致性描述的孪生网络模型,分析了AI内生的SDN切片生命周期；最后,给出了SDN切片架构下的虚拟网络功能(VNF)部署案例,并详细说明了AI内生优化框架和具体工作流程。实验结果表明面向白盒交换机的SDN切片测试平台实现了物理网络的拓扑可视化与流量全息,降低了网络的时延和带宽开销；且相较于非数字孪生AI算法,数字孪生深度确定性策略梯度(DT-DDPG)算法在增强型移动带宽(eMBB)场景下的时延和带宽消耗分别降低了46%和37%。

Abstract:

With the rapid development of the power industry, people have more demand for electricity, and some important scientific research projects are also deepening in the direction of electricity, but the current power Internet of Things lacks an SDN slice test platform built by white-box switches. Firstly, a SDN slicing platform architecture that integrates digital twin (DT) and AI intelligence endogenous is designed. Secondly, through the unified characterization of the white-box switch and link data, the twin network element ontology model, link ontology model and twin network model with consistent description with the physical network are constructed, and the endogenous SDN slice life cycle of AI is analyzed. Finally, a virtual network function (VNF) deployment case under SDN slicing architecture is given, and the AI endogenous optimization framework and specific workflow are explained in detail. Experimental results show that the SDN slicing test platform for white-box switches realizes topology visualization and traffic holography of physical networks, reduces network latency and bandwidth overhead. Compared with non-digital twin AI algorithms, digital twin deep deterministic policy gradient (DT-DDPG) algorithms reduce latency and bandwidth consumption by 46% and 37%, respectively, in enhanced mobile bandwidth (eMBB) scenarios.