Comparison
Event streamvsMessage queue
Event stream
consumers read from a retained, ordered log at their own position, so a new consumer can start from the beginning of history.
A durable, ordered, replayable log partitioned for parallelism, as opposed to a queue that deletes on acknowledgement. Retention means you can add a consumer, reprocess after a bug, or rebuild a projection. Ordering is only guaranteed within a partition, so the partition key decides what "in order" means for your data.
Full entry →Message queue
you push the slow job onto a queue and return to the user immediately, and a worker picks it up whenever it can.
A buffer of work items consumed by workers, where each message is delivered to one consumer and removed once acknowledged. It decouples request latency from work duration and absorbs bursts. It also converts synchronous errors into invisible ones, so a queue without a dead letter path and depth alerting is a place for work to go and quietly die.
Full entry →