The tech behind weather forecasting
summary
Buienradar runs on a pipeline that moves a 2008 monolith through microservices into Docker and Kubernetes with Argo CD and GitOps; seven engineers handle peaks of 15 million unique users and 5.8 billion pageviews per year from 80+ radars while sending millions of alerts daily. Key specifics: aggressive autoscaling (pods jumping 10→50 in seconds), canary releases, OpenTelemetry→Grafana/Prometheus observability, circuit breakers and retries, one P1 last year resolved in 30 minutes, and reported 100% availability at highest load.
AI-written summary · about
chapters (14)
- 0:00Weather forecasting tech intro
- 6:49Bioradar product intro
- 10:36Weather radar precipitation
- 13:04Monolith to microservices
- 16:15Docker containerizing services
- 18:24Kubernetes infrastructure transition
- 21:18GitOps with Argo CD
- 24:50System architecture overview
- 28:03DevOps reliability techniques
- 35:03Failure handling and incidents
- 37:53Autoscaling traffic spikes
- 40:43Spot instances cost optimization
- 42:51Future AI innovation plans
- 44:58Q&A
related talks
42mRobust Automated Forecasting in Python and R
31mFinding Needles in a Growing Haystack
12mYr - The next generation weather visualization
43mBuilding a community fountain around your data stream
08m24/7 Virtual Weather: Designing the Ultimate Weather Network
33mData: The Land DevOps Forgot
37mKeynote – Carol Willing
06mRuhrJS 2016 - Ben Schwarz