基于深度学习算法的SDN路由动态优化方法

Study on Dynamic Optimization Method of SDN Routing Based on Deep Learning Algorithm

传统的软件定义网络(software defined network,SDN)路由动态优化方法无法对路由节点数据进行动态分组,导致高数据转发场景中的网络时延抖动较长,因此,提出了基于深度学习算法的SDN路由动态优化方法。首先,根据SDN层次结构,动态规划最短传输路径;然后,通过计算链路生存时间,实现路由节点数据的动态分组;其次,基于分组结果、路由节点的匹配和转发特性,构建多目标路由动态优化模型;最后,利用深度学习算法建立网络信道参数与动态优化解之间的映射关系,从而实现SDN路由动态优化。仿真实验结果表明,该方法能够抑制网络时延抖动,降低网络时延。

Traditional software defined network(SDN)routing dynamic optimization methods can′t dynamically group routing node data results in longer network delay jitter in high data forwarding scenarios.Therefore,a dynamic optimization method for SDN routing based on deep learning algorithms is proposed.Firstly,according to the SDN hierarchy,the shortest path of transmission is dynamically planned.Then,by calculating the link lifetime,the dynamic grouping of routing node data is achieved.Secondly,based on the grouping results and matching and forwarding characteristics of routing nodes,a multi-objective routing dynamic optimization model is constructed.Finally,deep learning algorithms are used to establish the mapping relationship between network channel parameters and dynamic optimization solutions,so as to achieve dynamic optimization of SDN routing.The simulation experimental results show that this method can suppress network delay jitter and reduce network delay.