A conference talk transcript exploring the question of the optimal number of threads in Java applications. Covers CPU-bound vs IO-bound thread pool sizing formulas, benchmarking results on an 8-core system, Java 21 virtual threads and their limitations, Kubernetes millicores, and the ultra-low-latency single-threaded approach used at financial exchanges like LMAX. Concludes with a pragmatic argument that multi-threaded code is hard to read, test, debug, and maintain, so developers should minimize complexity where possible.

13m watch time

Sort: