基于度中心性的AFDX网络拓扑生成

航空电子系统随着任务需求和技术的发展不断向深度综合演进，其系统的复杂性给网络的设计和验证带来了巨大的挑战，如何通过网络生成实现受限资源条件下航电信息交互的实时性能保障是目前亟待解决的问题。针对目前存在的无法对航电网络进行实时性调控的拓扑设计方法进行改进，依据终端节点之间所有虚拟链路的最大通信帧长之和的大小关系，提出一种基于度中心性理论的航空电子全双工交换式以太网（AFDX）网络拓扑生成算法。将终端节点之间数据帧长作为节点度的衡量标准，对所有终端节点进行集合划分，并根据集合中终端节点的数据帧长对交换机进行连接。采用确定性网络演算以及仿真的方法对基于度中心性的AFDX网络拓扑生成算法进行效能评估。利用确定性网络演算方法，在小规模虚拟链路（VL）的组网下，结果显示:基于度中心性的拓扑生成算法生成的网络拓扑中75%的VLs实时性能优于原始人工设计的网络拓扑，且端到端延迟平均减小9.37%。利用OMNet++仿真方法，在1 400条虚拟链路的组网规模下，结果显示:基于度中心性的拓扑生成算法生成的网络拓扑中94.3%的VLs实时性能优于人为规划网络拓扑，且端到端延迟平均减小50.2%。由此表明:基于度中心性的拓扑生成算法很大程度上提高了网络的实时性能保障。 

AFDX network topology generation based on degree centrality

With the development of mission requirements and technologies, avionics systems continue to evolve into deep integration, and the complexity of systems has brought enormous challenges to the design and verification of networks. How to realize real-time performance guarantee of avionics information interaction through network generation under restricted resources is an urgent problem to be solved. According to the relationship between the sum of the maximum communication frame lengths of all virtual links between terminal nodes, the avionics full duplex switched Ethernet (AFDX) network topology generation algorithm based on degree centrality theory is proposed to improve the existing topology design method that cannot control the avionics network in real time. All the terminal nodes are collectively divided according to the data frame length between the terminal nodes which is used as a measure of the degree of the node. The switch performs dynamic connection according to the data frame length of the terminal node in the set. Deterministic network calculus and simulation methods are used for performance evaluation of AFDX network topology generation algorithm based on degree centrality. The results show that 75% of the VLs' real-time performance in the network topology based on degree centrality is better than the original artificially designed network topology using the deterministic network calculus method under the networking of small-scale virtual link, and the end-to-end delay is reduced by on average of 9.37%. The results show that 94.3% of VLs real-time performance in the network topology based on degree centrality is better than the artificially planned network topology. And the end-to-end delay is reduced by 50.2% on average using the OMNet++ simulation method under the networking scale of 1 400 virtual links. Therefore, the results show that the topology generation algorithm based on degree centrality greatly improves the real-time performance guarantee of the network.