Distributed Video Management at Tesco Scale with Edge Compute and Kubernetes
summary
Built in-house, a modular multi-stage video pipeline ingests RTSP via ONVIF, fronts streams with a serve stage so one camera supports many clients, records to hybrid storage and plays back over HLS while live uses WebRTC for sub-second latency. Edge compute runs on k3s Kubernetes per store, uses edge filtering so cloud pulls are on demand, and Prometheus-style observability controls cost. The deployment claims management of 200,000+ cameras, 4.8M recording hours/day (~4.5 PB/day) and real-time alerts, and it operates at scale.
AI-written summary · about
chapters (16)
- 0:00Introduction to Tesco
- 2:12Tesco video scale challenge
- 3:36Video use cases at Tesco
- 4:39Why build a video platform
- 6:46Demo of video platform UI
- 8:35Architecture constraints and challenges
- 11:02Solving camera fragmentation
- 12:47Video pipeline architecture
- 14:34Edge compute hosting choice
- 15:59Distributed storage at edge
- 16:40Kubernetes at the edge
- 17:24Hosting environment fragmentation
- 18:05Cloud-edge federation trade-offs
- 19:32Observability at scale
- 20:16Cost and data balance
- 20:37Key learnings and conclusion
related talks
25mEdge-to-Core Machine Learning and Data Pipeline on k8s
36mEdge Computing and Kubernetes BOF
40mCase Study: Kubernetes at Comcast
36mEdge Computing in Kubernetes: Best Practices for the Future
32mKubeEdge DeepDive: Extending Kubernetes To the Edge With Real-World Industry Use Case
42mDelivering TV Everywhere with Cloud Native Solutions
41mKubeEdge
07mData Gravity and Kubernetes: Managing Large-Scale Data Ingest With Minimal Latency