PreConfig: A Pretrained Model for Automating Network Configuration

TL;DR

PreConfig addresses the inflexibility of manual NCA tools by reframing network configuration tasks as text-to-text transformations that a pretrained LM can unify across generation, translation, and analysis. The authors construct an in-domain corpus, employ LLM-based data augmentation, and deploy a ConfigExtract-driven data mining pipeline to enable multi-task fine-tuning. Experiments show that PreConfig achieves strong performance on configuration generation (BLEU $82.25$, ROUGE $95.76$, EM $40.0$), analysis (BLEU $63.95$, ROUGE $93.35$, EM $10$), and translation (BLEU $70.84$, ROUGE $89.2$, EM $25$), outperforming generic LMs and existing tools. This work demonstrates a practical, scalable path for automated network configuration with domain-adapted pretrained models and multi-task learning.

Abstract

Manual network configuration automation (NCA) tools face significant challenges in versatility and flexibility due to their reliance on extensive domain expertise and manual design, limiting their adaptability to diverse scenarios and complex application needs. This paper introduces PreConfig, an innovative NCA tool that leverages a pretrained language model for automating network configuration tasks. PreConfig is designed to address the complexity and variety of NCA tasks by framing them as text-to-text transformation problems, thus unifying the tasks of configuration generation, translation, and analysis under a single, versatile model. Our approach overcomes existing tools' limitations by utilizing advances in natural language processing to automatically comprehend and generate network configurations without extensive manual re-engineering. We confront the challenges of integrating domain-specific knowledge into pretrained models and the scarcity of supervision data in the network configuration field. Our solution involves constructing a specialized corpus and further pretraining on network configuration data, coupled with a novel data mining technique for generating task supervision data. The proposed model demonstrates robustness in configuration generation, translation, and analysis, outperforming conventional tools in handling complex networking environments. The experimental results validate the effectiveness of PreConfig, establishing a new direction for automating network configuration tasks with pretrained language models.

Figures (12)

• Figure 1: An example of text transformation for configuration tasks. Configuration generation and analysis tasks involve the transformation between network configuration and natural language, while the configuration translation task involves the transformation of configurations from different vendors.
• Figure 2: (a) is an example network, where A' represents a Juniper router used as a backup for Cisco router A. (b) is an example configuration intent for network operators, used to configure route-maps for router A.
• Figure 3: Cisco and Juniper configuration fragment of route maps. The parts highlighted in the same color in (a) and (b) represent configuration modules with the same functionalities from different vendors.
• Figure 4: Overall framework of PreConfig. It consists of two stages. In stage 1, the pretraining data is collected through task-related corpus selection and LLM-based data augmentation. The pretrained model is obtained through continuous pretraining on a programming language model. In Stage 2, task data mining is accomplished through an LLM-based agent. The model acquires the capability to handle multiple tasks through multi-task learning based fine-tuning.
• Figure 5: Implementation of configuration data augmentation utilizing prompt engineering. The inputs to GPT are configuration snippets, prompt templates, and the outputs of GPT are expanded configuration data.