Multi-vendor interoperability and disaggregation are attracting the interest of network operators as a way to avoid vendor lock-in, thus opening the market and, possibly, reducing costs. NETCONF, in concert with YANG data models, has been identified as the software defined networking (SDN) configuration protocol for these networks. Currently, several YANG models describe devices in a vendor-neutral way; however, there is a lack of YANG models describing functions. Moreover, the centralization of the control plane may suffer from scalability issues during critical situations (e.g., link failures), given that several restoration requests will arrive at the SDN controller close in time. This paper investigates an innovative paradigm of restoration for disaggregated SDN networks named delegated restoration, which operates in a hybrid centralized–distributed manner. Before a failure occurs, the backup lightpath is centrally computed by the SDN controller, and, based on this computation, the controller informs the network devices (switches and transponders), through NETCONF, of the reconfigurations to perform in case of failure. Simulations show that delegated restoration can reduce the restoration time with respect to a fully centralized approach, making it a candidate for impacting the operation, administration, and maintenance of next-generation networks.

中文翻译：

---

SDN分解网络中的本地自动恢复

多供应商的互操作性和分解正在吸引网络运营商的兴趣，以此避免供应商陷入困境，从而打开了市场并可能降低了成本。NETCONF已与YANG数据模型一起被确定为这些网络的软件定义网络（SDN）配置协议。当前，YANG的几种模型以与供应商无关的方式描述设备。但是，缺少描述功能的YANG模型。此外，鉴于几个恢复请求将及时关闭到达SDN控制器，因此在关键情况下（例如，链路故障），控制平面的集中化可能会遇到可伸缩性问题。本文研究了分布式SDN网络的一种创新还原模式，即委托还原，它以集中式-分布式混合方式运行。在发生故障之前，备用光路由SDN控制器集中计算，并且基于此计算，控制器通过NETCONF通知网络设备（交换机和转发器）有关重新配置的信息，以便在发生故障时执行重新配置。仿真表明，相对于完全集中的方法，委托还原可以减少还原时间，使其成为影响下一代网络的运行，管理和维护的候选对象。