A Social Network Under Social Distancing: Risk-Driven Backbone Management During COVID-19
summary
Facebook describes a production Risk Simulation System (RSS) that runs year-long failure simulations to quantify backbone risk, modeled with non‑exponential, multimodal subsea repair times, and implemented in 18,000 lines of C++. RSS runs fine‑grained scenarios in about 250 seconds and core‑score cases in ~0.1s, detects rising demand loss during COVID, drives actions like capacity additions and downgrading QS class 2 traffic to classes 3–4, and shows traffic negatively correlates with mobility.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:45Backbone Network
- 2:27Network Risk
- 2:55Unified Risk Metrics
- 3:45Risk Simulation System
- 4:47Finegrained Risk Simulation
- 6:25Failure Modeling
- 7:33Stress Test
- 8:37QOS Assignment
- 9:35Insights
- 11:20Traffic Forecast
- 12:32Conclusion
related talks
46mLessons learned from monitoring a large production network
19mEngineering Facebook's Data Center Network: Challenges and Limitations
38mFacebook's Data Center Fabric
32mActive Fault Finding in Networks
39mMobile Networking Challenges - Networking @Scale
34mHow We Drained Every Backbone Router Simultaneously
21mFrom Reactive to Proactive: How AI is Redefining Resilience in Modern Networks
39mLoad Balancing Building Blocks