コレオグラフィとオーケストレーションの選択

You have met both saga styles. Now the practical question: which one should you use? The choice shapes coupling, visibility, and operability of your whole workflow.

Choreography: services react to each other's events, no central brain. Orchestration: a central orchestrator explicitly tells each service what to do next.

Choreography couples services only to event contracts — very loose. Orchestration centralizes flow knowledge in the orchestrator, coupling services to it but freeing them from knowing about each other.

With orchestration the entire workflow lives in one place — easy to read, trace, and reason about. With choreography the flow is emergent, scattered across event handlers, and harder to follow.

Tracing a failure is simpler in orchestration: the orchestrator logs each step. In choreography you must reconstruct the path from distributed event logs, which gets painful as steps grow.

Adding a step in choreography can mean changing several services that emit/consume events. In orchestration you usually change only the orchestrator, leaving participants untouched.

Choreography can drift into hard-to-spot event cycles where services trigger each other in loops. Orchestration's explicit flow makes such cycles obvious and avoidable.

The orchestrator is a potential bottleneck and single point of failure (mitigated by making it stateless and replicated). Choreography has no such central component.

Practical guidance:

• Few steps, simple flow → choreography
• Many steps, complex branching, needs visibility → orchestration

Real systems often mix both: orchestrate the complex core workflow, while letting peripheral concerns (analytics, notifications) react to events choreographically. Pick per workflow, not per system.

Conway's law applies: independent teams may prefer choreography to avoid a shared orchestrator. A platform team that owns the workflow may prefer the central control of orchestration.

You can now choose a saga style:

• Choreography = loose coupling, emergent flow; orchestration = central control, clear visibility
• Orchestration eases debugging, adding steps, and avoiding cycles
• It introduces a central component to keep resilient
• Few simple steps favor choreography; many complex steps favor orchestration
• Hybrids are common — decide per workflow