CableMon: Improving the Reliability of Cable Broadband Networks via Proactive Network
summary
Proposes CableMon, a public ML system that turns eight months of cable PNM time-series and customer trouble tickets into statistical features and automated thresholds, using moving averages, deltas, variance and a ticket-driven threshold search that maximizes a ticketing-rate ratio. Evaluated on ISP data, detected anomalies overlap 81.9% with at least one ticket (four times better than public tools) and account for 23.0% of network-related tickets, enabling proactive sliding-window dispatch from four-hour PM samples.
AI-written summary · about
chapters (15)
- 0:00Intro
- 1:02Cable Networks are Unreliable
- 2:19Roadmap
- 3:01Proactive Network Maintenance (PNM)
- 4:01Existing Work
- 7:01Datasets: PNM Data and Customer Trouble Tickets
- 8:11Tickets Correlate with Network Faults
- 9:11Challenges
- 10:08Filter Non-network Related Tickets
- 11:30Determine the Threshold
- 14:16Selecting Top Features
- 14:42Top Five Features
- 14:56Combining Features
- 16:12Evaluation
- 18:44Conclusion
related talks
16mClosed-loop Network Performance Monitoring and Diagnosis with SpiderMon
17mCharacterizing Physical-Layer Transmission Errors in Cable Broadband Networks
36mNetwork Machinery: A United-Front for Network Troubleshooting with CRDs
46mLessons learned from monitoring a large production network
14mSkyline: A Cloud Centric Internet Monitoring Engine
16mEnhancing Global Network Monitoring with Magnifier
35mNetwork Management and Monitoring with Python
24mHarp: Improving VPC Network Availability via Efficient Failure Detection and Rerouting...