Opportunities and Implementation of Neural Machine Translation for Network Configuration

Abstract

Network configuration translation is proposed to solve the problem of configuration migration and backup in the process of network configuration management, and how to automate configuration translation is an important issue in the development of autonomous driving network. It is of great scientific significance and application value to intellectualize network configuration management via artificial intelligence technology represented by machine learning. Recently, neural machine translation (NMT) technology has been widely used in natural languages and programming languages to integrate and translate heterogeneous languages, which provides new insights into using machine translation to tackle the problems in network operation and maintenance. In this article, we analyze the possibilities and challenges involved in the application of NMT in network configuration, propose a general framework of the network configuration NMT model, and report on experiments with an implementation of a configuration translation use case that takes an unsupervised approach. The design, construction and deployment of the model are explained in detail and the feasibility is validated experimentally. We also discuss the major challenges and potential research directions of NMT in network configuration.