Scaling IoT
summary
Quby builds the tone thermostat, runs 175,000 connected units and projects 350,000 this year but needs 1.5–2 million devices to be profitable. The deployed stack moves from private datacenter + heavy VPN concentrators to Amazon, using Docker and Marathon for app lifecycle, MQTT brokers (HyQ proved fast but costly), RabbitMQ, ElasticSearch/Kibana/Fluentd for logs, and a CI/CD + infrastructure-as-code pipeline. Engineering hits include a 1.2 million-connection broker test, 50% datacenter load on VPN, and cutting environment provisioning from ~20 minutes/3 weeks to ~4 minutes.
AI-written summary · about
chapters (10)
- 0:00Introduction and company background
- 2:37Product evolution
- 6:42Technical architecture
- 9:23Growth and challenges
- 11:46Infrastructure transformation
- 20:04DevOps and monitoring
- 22:32Multi-tenant strategy
- 26:15Connectivity and scaling
- 36:56Lessons and conclusions
- 40:15Q&A session
related talks
16mBuilding scalable IoT systems with MQTT and AWS
23mKeynote: Intelligent Internet of Things: Start Analyzing Your Global Device Data for Real-Time
46mBuild, Scale, & Create IoT and Mobile connected applications with the Cloud
31mFun with Sensors
27mIoT is hard, keep hustling
47mLessons and Advances in IoT: Toward 1M Connected Devices
45mIoT End to End
55mIoT Solution Recipes - Beyond your first slice of Raspberry Pi