Building Production-Ready ML-Based Network Routing
summary
Claims concrete wins: a production ML routing stack with telemetry pipeline, model training and real‑time inference plus protocol fallbacks and rollbacks that cut latency spikes by almost 70% and saved millions in one case and reduced page load time by 30% and manual interventions by 95% in another. Architecture uses BGP/SPF as backups, encrypted telemetry, anomaly detection, and dashboards for drift and confidence. Deployment is phased: shadow → advisory → selective → full automation, and models use supervised, reinforcement or a hybrid training approach.
AI-written summary · about
chapters (14)
- 0:00Introduction and Overview
- 0:29Traditional vs. ML-Based Routing
- 1:45Technical Architecture of ML-Based Routing
- 2:28Comparing ML and Traditional Routing
- 2:54Data Collection and Model Training
- 3:40Deployment Phases of ML Routing
- 4:28Monitoring and Observability
- 5:01Security Considerations
- 5:30Operational Considerations
- 6:00Real-World Case Studies
- 6:41ROI and Risk Management
- 7:13Implementation Roadmap and Best Practices
- 7:34Team Structure and Skills
- 7:54Conclusion and Future Outlook
related talks
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
09mProduction-Ready MLOps for Telecommunications
19mTowards Domain-Specific Network Transport for Distributed DNN Training
14mPerformance Optimization in High-Volume Transaction Systems
21mFrom Reactive to Proactive: How AI is Redefining Resilience in Modern Networks
25mContinuous ML Applications in Production
40mBuilding highload ML powered service
14mClosing the Train-Inference Divide for Asynchronous RL at Scale