ExactlyonceSemantik

ExactlyonceSemantik is a concept in distributed systems describing guarantees that a given operation or message is processed exactly once. In practice this means the system avoids duplicates and ensures side effects occur only once. It is contrasted with at-least-once (possible duplicates) and at-most-once (possible losses).

Achieving EOS relies on coordination, reliable storage, and idempotent design. Techniques include idempotent operations, use of unique processing identifiers (sequence or correlation IDs) to detect duplicates, and transactions that atomically commit related steps. In messaging and streaming, EOS is often implemented with idempotent producers and transactional writes.

However, guaranteeing EOS across external effects is difficult; outside interactions or long-running processes may require compensating actions. Trade-offs include performance overhead and design complexity.

Common patterns include outbox tables, the Saga pattern, two-phase commit where applicable, and ensuring idempotency across services.

Exactly-once semantics are critical in financial transactions, inventory updates, and order processing. The term is widely used in technical literature, including German-language contexts as ExactlyonceSemantik.