distributed-systems-patterns

You need a saga pattern when a transaction spans multiple services and requires coordinated compensations to maintain data consistency. It applies to event-driven architectures where distributed transactions are necessary and explicit consistency trade-offs must be defined.

Choose orchestration vs choreography by evaluating workflow complexity and coupling. Orchestration centralizes coordination for explicit control, while choreography uses event-driven reactions for decentralized autonomy, with decision criteria and step-by-step reasoning provided for selection.

The outbox pattern ensures reliable event publishing within the same database transaction as state changes. You should avoid it if your system cannot handle the eventual consistency trade-offs or lacks the infrastructure to reliably poll and publish transactional outbox entries.

Implement idempotency strategies by defining unique identifiers for events and tracking processed messages to prevent duplicate execution. This ensures reliable cross-service workflows by making operations safely retryable without side effects.

CQRS works with event sourcing considerations by separating read and write models to optimize multi-service workflows. It satisfies requirements for explicit consistency trade-offs, though it should be evaluated against your specific query latency and scaling needs.