Choreographed microservices talk to each other asynchronously, blindly broadcasting notifications into a service cloud. Those notifications are handled by whatever client services are interested. These systems eliminate many of the problems associated with orchestrated systems (which work more like synchronous function calls) and are typically much faster than orchestrated systems, but they have their own idiosyncrasies and implementation challenges. Allen Holub explores the inherent problems in orchestrated systems and then looks at how choreography can solve those problems. Allen explores three approaches to choreography: HTTP based, pub/sub messaging based, and brokerless swarming systems. He introduces appropriate messaging architectures and frameworks and looks at several practical examples. Finally, Allen looks at event storming: one of the best approaches to designing choreographed systems. You’ll leave with an understanding of both why you should be using choreography for most high-level APIs and how to design and build these systems.

This session was recorded at the 2019 O'Reilly Software Architecture Conference in San Jose.