Production Tuning for Temporal Workers at Scale

A webinar covering how to tune and scale temporal workers in production. Marcus and Sam from temporal explain the core concepts of durable execution, how workers poll task queues and use sticky caching for performance, and then walk through the three main tuning knobs: compute (executor slots), memory (sticky workflow cache), and IO (task pollers). Key advice includes applying the theory of constraints to identify bottlenecks before scaling, since blindly adding workers can worsen problems like downstream API rate limits. Language-specific considerations are noted for Java (max workflow thread count) and TypeScript (V8 sandbox options).