Finding Network Misconfigurations by Automatic Template Inference
summary
Automatic inference groups similar router config snippets into parameterized templates and flags small groups as outliers; implemented in the tool SelfStarter and evaluated on Microsoft datacenter, Microsoft WAN, and a large campus network. Templates use block-aware sequence alignment and minimum-cost bipartite matching to allow safe reordering and parameter abstraction. Output is a concise meta-template that highlights exact line-level differences and suggested fixes. On real networks this detects 43 previously unknown bugs and surfaces ordering, missing-line, and spurious-community errors without any formal spec.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:20Network Misconfigurations Are Common
- 0:41Network Verification
- 1:21Limitation: Lack of Specifications
- 1:56PROBLEM: How to find misconfigurations without an explicit specification?
- 2:44Network Device Roles
- 3:12PROBLEM: How to model config "similarity and define "deviance ?
- 4:49Contributions
- 5:43End-to-End Design
- 7:29An Example: Campus Network
- 8:50SelfStarter Example Output
- 9:58Group 1 Template
- 10:50Deviance Identification
- 12:24Template Inference: Key Ideas
- 16:06Results - Summary
- 17:50Results - Sources of Misconfigurations
- 19:10Conclusion
related talks
27mNetComplete: Practical Network-Wide Configuration Synthesis with Autocompletion
16mTest Coverage for Network Configurations
18mNetKeeper: Enhancing Network Resilience with Autonomous Network Configuration...
24mAutomatically Correcting Networks with NEAt
12mMetha: Network Verifiers Need To Be Correct Too!
30mUsing Python to design, configure, and measure large-scale networks
27mJuggling the Jigsaw: Towards Automated Problem Inference from Network Trouble Tickets
16mConfig-Snob: Tuning for the Best Configurations of Networking Protocol Stack