SysPad › Examples › Video streaming
Video streaming
Netflix-style: a video CDN over S3 origin + Glacier archive, a catalog API, a continue-watching cache, and a playback-events analytics pipeline.
Network-dominated cost (multi-MB CDN objects) and a playback-analytics pipeline.
- Load: 8,000 req/sec
- First to saturate: Auth
- Estimated cost: $1,636k/mo
How it works
- Continue-Watching
- Caches each viewer resume position so playback picks up instantly.
- Playback Events
- Playback pings stream out fire-and-forget; the whole analytics pipeline runs off the response.
Request flows
Start playback
- DNS
- API LB
- Streaming API
- Auth
- Continue-Watching
- Catalog
Playback heartbeat
- DNS
- API LB
- Streaming API
- Playback Events
- Event Aggregator
- Analytics Warehouse
Lessons for this architecture
- How Route 53 works
- How CloudFront works
- How S3 works
- How S3 Glacier works
- How Application Load Balancer works
- How ECS Fargate works
- How Cognito works
- How Amazon ElastiCache works
- How DynamoDB works
- How Kinesis Data Streams works
- How AWS Lambda works
- How Redshift works
Worked step by step
The same interview question, built up one twist at a time, with a simulation at each stage.