A message queue or event-streaming platform lets one component publish work or events for another component to process asynchronously.

Example: order processing

• Order API validates and records an order.
• A message is published for downstream processing.
• A consumer sends confirmation email.
• Other consumers may update analytics or notify a warehouse.

Common technologies

• RabbitMQ: message broker with queueing and routing features.
• Apache Kafka: distributed event-streaming platform with durable streams and consumer groups.
• Amazon SQS: managed queue service.
• Delivery, ordering, retention, and operations differ by technology.

Important concepts

• Producer publishes messages; consumer processes them; broker distributes them.
• Acknowledgment records processing progress according to the protocol.
• Dead-letter queue stores messages that cannot be processed under the configured policy.
• Idempotent consumers handle duplicate delivery without repeating unintended effects.

Delivery semantics

• At-most-once: a message may be lost, but is not redelivered under the guarantee.
• At-least-once: a message may be delivered more than once.
• Exactly-once behavior is limited to defined system boundaries and does not automatically guarantee exactly-once effects in every external system