Uploaded January 2026 | Updated September 2026, 2 weeks ago

---

Can your Java application handle 270,000 requests per second without crashing?

Hugo Marques, a Software Engineer at Netflix, shares a "from the trenches" look at the evolution of a high-scale data processing job. Moving beyond theoretical documentation, Hugo breaks down the messy reality of optimizing for IO-bound problems at massive scale.

⏱️ Video Timestamps (For Navigation)
0:00 - Introduction: Why "The Trenches"?
1:45 - The Scale: 270K RPS and IO-Bound Problems
3:20 - Toolbox: Threads, CompletableFuture, and Virtual Threads
5:50 - The First Mistake: Parallel Stream Greed
9:10 - Solving CPU-Bound vs. IO-Bound Issues
11:30 - First Out of Memory: Chaining Async Tasks
14:45 - Protecting Dependencies: Rate Limiters & Semaphores
18:20 - Virtual Threads & The "Pinning Problem"
22:10 - Simplifying the Solution: Bounded Executors
25:40 - Results: Performance Side-by-Side
28:15 - Key Takeaways for Senior Engineers
30:50 - Q&A: Sharding Load & Handling Retries

🔗 Transcript available on InfoQ: bit.ly/45yE3jf

#Java #Netflix #SoftwareArchitecture #VirtualThreads #BackendEngineering #InfoQ