Cost vs Stability in a Cloud Environment
summary
Nubank runs ~2,000–4,000 Clojure microservices, ~200 deploys/day, backing a 100M+ customer base and massive PIX traffic that forces rethinking scale. Engineering choices include moving from G1 to ZGC to avoid stop‑the‑world pauses that trigger needless autoscaling, using AWS Spot instances plus NVMe disks to shift cache hits from external stores (spending $1 on NVMe avoids ~$3,500 in scaling), and adopting CloudZero for cost observability so business units own buying decisions and stability becomes the cost-optimization.
AI-written summary · about
chapters (9)
- 0:00Intro
- 5:55Pix payments in Brazil
- 11:52Cost vs Stability
- 15:20Garbage collector drama
- 21:31Culture change
- 26:39Stable efficiency
- 27:15What's next?
- 27:32Continuous resilience
- 28:16See you next year?
related talks
31mKeeping Your Cloud Footprint in Check
39mHow automating cloud operations accelerated time to market at NU
39mCloud Native at PostFinance - Is it completely normal?
48mCloud Chaos & Microservices Mayhem
36mMigrating Applications to the Cloud in a Cost Effective Manner
39mThe Power of Cloud Native in Financial Institutions
27mCloud Native: why, oh why?
18mKeynote: Mind the Gap: Bridging Cloud Native Innovation with Real-World Use