# Fault Localization in Large-Scale Network Policy Deployment

**Authors:** Praveen Tammana, Chandra Nagarajan, Pavan Mamillapalli, Ramana Rao Kompella, Myungjin Lee  
**arXiv:** [1712.08129](https://arxiv.org/abs/1712.08129)  
**Date:** 2017-12-21  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1712.08129v1/

## Key points

- Recent advancements in network management automation and Software-Defined Networking (SDN) simplify network policy management tasks
- Network policies are represented in an abstract model at a centralized controller and deployed as low-level rules across various network devices
- Any software or hardware component within this cycle can cause underlying network issues
- The paper titled "Fault Localization in Large-Scale Network Policy Deployment" addresses the problem of network policy fault localization
- The authors propose a solution through risk modeling and a greedy algorithm to identify faulty policy objects within the network policy
- They design and develop SCOUT, a fully automated system that identifies faulty policy objects and pinpoints physical-level failures that caused them
- Evaluation of SCOUT involves real testbed experiments and extensive simulations, demonstrating its ability to detect faulty objects with minimal false positives and false negatives
- This research highlights the importance of addressing fault localization in large-scale network policy deployment
- The authors contribute to improving the reliability and performance of network management frameworks.

## Summary

The recent advancements in network management automation and Software-Defined Networking (SDN) have greatly simplified network policy management tasks. However, these new technologies also introduce a new mode of failure within the management cycle itself. Network policies are represented in an abstract model at a centralized controller and then deployed as low-level rules across various network devices. As a result, any software or hardware component within this cycle has the potential to cause underlying network issues. In this paper titled "Fault Localization in Large-Scale Network Policy Deployment," authors Praveen Tammana, Chandra Nagarajan, Pavan Mamillapalli, Ramana Rao Kompella, and Myungjin Lee address the problem of network policy fault localization that arises in operating policy management frameworks for production networks. They propose a solution by formulating the problem through risk modeling and introducing a greedy algorithm that efficiently identifies faulty policy objects within the network policy. To further enhance their solution, the authors design and develop SCOUT, a fully-automated system that not only identifies faulty policy objects but also pinpoints physical-level failures that caused these objects to become faulty. The evaluation of SCOUT involves real testbed experiments and extensive simulations, which demonstrate its ability to detect faulty objects with minimal false positives and false negatives. Overall, this research highlights the importance of addressing fault localization in large-scale network policy deployment. By presenting an effective solution through risk modeling and algorithmic techniques, the authors contribute to improving the reliability and performance of network management frameworks.