Design a High-Throughput Smart Home Telemetry Ingestion System

1. Requirements

• Functional: Accept streaming telemetry (temperature, status) from 10 million connected devices (e.g., hybrid Wiz, Xiaomi, and custom setups).
• Non-Functional: 99.99% availability, latency under 200ms for critical alerts.

2. Component Breakdown

• Edge Gateway: Handles local MQTT termination to reduce cloud round-trips.
• Ingestion Layer: Load-balanced WebSocket servers feeding into a Kafka cluster.
• Storage: Time-Series Database (TSDB) for metrics, Document DB for device states.

3. Tradeoffs & Bottlenecks

• Using MQTT over WebSockets reduces header overhead, but requires keeping millions of stateful connections open at the load balancer level.