Revamping How Node.js Binaries Are Served At Scale
summary
Node release assets now route through Cloudflare Workers + R2, replacing a single DigitalOcean origin and nginx with serverless edge logic. The system handles ~2.4 billion requests and 2.4 PB/month, with worker CPU time around 0.85 ms, P50 latency ~25 ms, 250 req/s sustained and 0% error rate since rollout. Four rewrites produce a middleware router, Sentry for observability, an R2 client swap, and planned KV caching for directory listings.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:03Talk purpose and infrastructure
- 2:55Traffic statistics
- 5:25Original setup problems
- 7:13Short-term fixes
- 9:19Long-term requirements
- 10:46Why Cloudflare Workers
- 12:30New infrastructure design
- 14:13Implementation challenges
- 16:39Debugging complexities
- 17:43Monitoring improvements
- 18:46Performance metrics
- 19:50Future plans
- 21:17Conclusion
related talks
42mFrom Localhost to Production
25mFrom PHP and on-prem servers to NodeJS/React and AWS
32mPushing boundaries to the edge
31mNode.js that's hugely reliable, fast, and scalable
31mKeynote Panel: Node.js at Massive Scale
26mKeynote talk - The future of Node.js
17mBuilding a High Performance, Dynamic CDN with Next.js
46mNodeJS Scalibility Tips