CEGS: Configuration Example Generalizing Synthesizer
summary
CEGS is a system that uses a GNN-based Querier to find relevant device documentation examples, a GNN-based Classifier to map example snippets to target devices, and an LLM-driven synthesis pipeline (GPT‑4) that generates templates, runs syntax + local + global verifiers, and fills parameters with a formal synthesizer. It automatically produces correct configs without human intervention for networks up to 1,094 devices across ~100 scenarios, and runs over 30× faster than prior LLM-based approaches that take hours.
AI-written summary · about
chapters (10)
- 0:00CEGS synthesizer intro
- 1:45Limitations of synthesizers
- 3:10Key automation challenges
- 5:00CEGS phases
- 5:21Phase one: retrieval
- 7:09Phase two: associations
- 8:39Phase three: generation
- 11:10CEGS performance
- 15:47Contributions and results
- 16:30Q&A
related talks
27mNetComplete: Practical Network-Wide Configuration Synthesis with Autocompletion
12mMetha: Network Verifiers Need To Be Correct Too!
15mA carrier approach to efficient CPE management
19mComputing Precise Control Interface Specifications
16mSirius: Composing Network Function Chains into P4-Capable Edge Gateways
13mSNI5GECT: A Practical Approach to Inject aNRchy into 5G NR
27mAlembic: Automated Model Inference for Stateful Network Functions
16mCetus: Releasing P4 Programmers from the Chore of Trial and Error Compiling