Software Defined Networking (SDN) decouples the control and data planes of a network, and employs a central controller to provide effective use of network resources and ease of service provision. Virtualization in SDN empowers the virtual SDN networks to share common physical network infrastructures, and, thus, make them capable for versatile service provisioning. In such setting, SDN hypervisor supports multiple virtual SDN network (vSDN) in which each vSDN has its own controller. To create virtual networks and achieving optimal sharing of physical network resources, virtual network embedding (VNE) algorithms are designed. Single domain VNE is a well-studied problem in NV literature. However, in most of the real life scenarios, VNs are provisioned across heterogeneous administrative domains belonging to different Infrastructure Providers (InPs).

In this paper, we propose a VNE algorithm, namely vSDN-CLA, using Irregular Cellular Learning Automata (ICLA) for multi-domain SDN networks. We consider two aspects – optimal mapping of virtual nodes and links into multi-domain substrate network, and optimal placement of SDN controller with respect to throughput and end-to-end delay. We extend the vSDN-CLA by considering dynamic flow migration to achieve resource optimal routing. We evaluate the proposed schemes using Mininet. We observe that the proposed schemes outperform the existing benchmark schemes with respect to throughput, end-to-end delay, and virtual network request acceptance ratio under both the single and multi-domain environments.

软件定义网络（SDN）将网络的控制平面和数据平面解耦，并使用中央控制器来提供对网络资源的有效利用并简化服务提供。SDN中的虚拟化使虚拟SDN网络能够共享常见的物理网络基础结构，从而使它们能够进行通用的服务供应。在这种设置下，SDN虚拟机管理程序支持多个虚拟SDN网络（vSDN），其中每个vSDN都有其自己的控制器。为了创建虚拟网络并实现物理网络资源的最佳共享，设计了虚拟网络嵌入（VNE）算法。在NV文献中，单域VNE是一个经过充分研究的问题。但是，在大多数现实生活场景中，VN跨属于不同基础结构提供程序（InP）的异构管理域进行设置。

在本文中，我们针对多域SDN网络使用不规则蜂窝学习自动机（ICLA）提出了一种VNE算法，即vSDN-CLA。我们考虑两个方面–虚拟节点和链接到多域基础结构网络的最佳映射，以及SDN控制器在吞吐量和端到端延迟方面的最佳放置。我们通过考虑动态流迁移来扩展vSDN-CLA，以实现资源最佳路由。我们使用Mininet评估提出的方案。我们观察到，在单域和多域环境下，在吞吐量，端到端延迟和虚拟网络请求接受率方面，拟议的方案都优于现有的基准方案。