How does Kafka differ from a traditional message queue?

The main difference is what happens to an event after it is read. Kafka stores events in a partitioned, replicated log and does not delete them when they are consumed: consumers can read a topic's events as often as needed, and retention is configured per topic (by time or size). A consumer tracks its position as an offset per partition and can deliberately rewind to an old offset and re-consume data, for example after fixing a bug (replay). Consumers pull data from the brokers instead of having it pushed, so a slow consumer simply falls behind and catches up when it can. Topics are multi-producer and multi-subscriber, and are split into partitions: events with the same key go to the same partition and are read in the order they were written, and within a consumer group each partition is consumed by one consumer at a time. A common use of this retained, replayable log is event sourcing and audit trails.

Expected key concepts: log, retention, replay, offset, rewind, pull, partitions, consumer group, multi-subscriber, event sourcing