A message queue, implemented by systems such as RabbitMQ, Apache Kafka or cloud services, sits between data ingestion and the services that process data. When a fleet reports at once, messages pile up in the queue instead of overwhelming the database, and consumers drain them at their own pace. If a consumer crashes, messages wait; if load grows, more consumers are added.
Queues differ in guarantees. Some deliver each message at least once, so consumers must tolerate duplicates; log-style systems such as Kafka retain messages for a period and allow replay, which is valuable for reprocessing after a bug. Protocols such as AMQP define queue semantics, while an MQTT broker is a related but lighter device-facing component. See The IoT data pipeline.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.