How Netflix Handles 270K RPS - Java Concurrency Lessons

Learn how to handle massive scale Java applications through a real-world case study from Netflix's engineering team in this 50-minute conference talk. Discover the practical challenges and solutions involved in processing 270,000 requests per second without system crashes, moving beyond theoretical documentation to explore the messy reality of optimizing IO-bound problems at enterprise scale. Explore essential Java concurrency tools including threads, CompletableFuture, and virtual threads while understanding common pitfalls like parallel stream overuse and memory issues from chaining async tasks. Master advanced techniques for protecting dependencies using rate limiters and semaphores, tackle the virtual threads "pinning problem," and implement bounded executors to simplify complex solutions. Gain insights into CPU-bound versus IO-bound problem solving, examine performance comparisons through side-by-side results, and absorb key takeaways specifically designed for senior engineers. Benefit from practical Q&A coverage on sharding load distribution and retry handling strategies, all presented by Hugo Marques, a Software Engineer at Netflix, who shares direct experience from high-scale data processing job evolution.