Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about CEGS (Configuration Example Generalizing Synthesizer), a novel approach to automated network configuration synthesis presented at NSDI '25. Discover how this system addresses the limitations of existing network configuration synthesizers that require extensive human involvement in drafting templates or coding in domain-specific languages. Explore the innovative combination of graph neural networks (GNNs) and large language models (LLMs) that enables CEGS to automatically identify configuration examples from device documentation, generalize them to arbitrary network topologies, and synthesize configurations without human intervention. Understand the three key components: a GNN-based Querier for identifying relevant examples, a GNN-based Classifier for topology generalization, and an efficient LLM-driven synthesis method for generating compliant configurations. Examine the evaluation results demonstrating CEGS's ability to automatically synthesize correct configurations for networks with over 1,000 devices, achieving performance more than 30 times faster than state-of-the-art LLM-based synthesizers while eliminating the need for human experts in the loop.
