The Google Cloud powers your Philips Hue Lightbulbs
summary
Philips Hue migrates its cloud from App Engine to App Engine Flex and then to Google Kubernetes Engine to enable persistent WebSocket connections directly to bridges, cutting remote-command latency from seconds to near real time. The production cluster runs 25 microservices across about 600 instances on ~60 machines, serving millions of homes while a small team of 10 manages it. Observability yields about 10 GB/day logs, metrics in Datadog, GitLab pipelines for automated deploys and ~30 minute push-to-production times.
AI-written summary · about
chapters (18)
- 0:00Introduction and welcome
- 1:03Philips Hue and its vision
- 2:07Starting with App Engine
- 4:34Initial cloud solution challenges
- 6:19Philips Hue market success
- 7:26Need for faster responses
- 8:30Evolution to Kubernetes
- 9:58Current infrastructure scale
- 10:41Logging and debugging struggles
- 12:52Metrics and monitoring insights
- 14:18Microservices and team growth
- 16:43Scaling the development team
- 18:27Knowledge sharing strategies
- 20:35DevOps and tooling evolution
- 21:40Metrics vs logging importance
- 22:45Microservices pitfalls
- 23:27Scaling the team challenges
- 24:31Future plans and advice
related talks
52mIoT on Google Cloud Platform
20mCollision 2015 - Urs Holzle of Google with Bruce Upbin of Forbes
57mDebugging Java in Production on Google Cloud Platform
21mMoving into the cloud
11mAbstracting infrastructure away from developers
25mAn Introduction To Google Cloud Platform
35mExperience from the field - how customers migrated from Google Cloud IoT Core to Eclipse Hono
39mOops. I broke the Google, now what?