Kafka vs SQS

High-throughput distributed log versus managed message queue for event streaming.

Overview

What is being compared

Apache Kafka is a distributed event log for high-throughput, ordered, replayable streams. SQS is a fully managed simple queue for point-to-point work distribution.

Model — Distributed event log — Simple message queue
Delivery — Any consumer group reads every message — Point-to-point, one consumer per message
After consumption — Messages stay for the configured retention period — Messages are deleted
Replay — Any consumer group can replay from any offset — Not possible
Operations — You run it, or use MSK — Fully managed

How Kafka works

Producers append to a log. Each consumer group keeps its own offset, so several groups read the same messages independently and any of them can rewind.

How SQS works

Consumers compete for messages on one queue. A message is processed once, then deleted.

The same message, two exchanges

A Kafka consumer only moves its own offset forward; the message stays in the log. An SQS worker receives the message, which is hidden from other workers while it is processed, then deletes it.

> [!TIP] The log and the queue side by side, three services reading every order event, and the choice between the two play step by step in Visual Explanation below.