17‑Point Comparison: Kafka vs RabbitMQ vs ZeroMQ vs RocketMQ vs ActiveMQ

Below is a detailed 17‑point comparison of Kafka, RabbitMQ, ZeroMQ, RocketMQ, and ActiveMQ when used as message‑queue solutions.

1. Documentation

Kafka: moderate. Official book by the author and some online resources. RabbitMQ: abundant. Several good books and many online materials. ZeroMQ: scarce. No dedicated book; most resources are code examples and brief introductions. RocketMQ: scarce. No dedicated book; official docs are concise, but technical details are limited. ActiveMQ: abundant. No dedicated book, but many online resources.

2. Programming Languages

Kafka: Scala RabbitMQ: Erlang ZeroMQ: C RocketMQ: Java ActiveMQ: Java

3. Supported Protocols

Kafka: proprietary protocol based on TCP RabbitMQ: AMQP ZeroMQ: TCP, UDP RocketMQ: proprietary protocol ActiveMQ: OpenWire, STOMP, REST, XMPP, AMQP

4. Message Storage

Kafka: memory, disk, database; supports massive accumulation. The smallest storage unit is a partition; topics contain multiple partitions distributed across brokers for load balancing and high availability. RabbitMQ: memory, disk; supports limited accumulation. Supports persistent and non‑persistent messages, with optional in‑memory backup for persistent messages. ZeroMQ: stored in sender’s memory or disk; no persistence. RocketMQ: disk; supports large accumulation. Uses CommitLog for data and ConsumeQueue as an index, enabling sequential writes and high write performance. ActiveMQ: memory, disk, database; supports limited accumulation.

5. Message Transactions

Kafka: supported RabbitMQ: supported (client channel set to transaction mode) ZeroMQ: not supported RocketMQ: supported ActiveMQ: supported

6. Load Balancing

Kafka: supports load balancing via partition leaders distributed across brokers; consumers in a group share partitions evenly; automatic rebalancing occurs when consumers join or leave. RabbitMQ: limited support; routing determined by exchange and routing key; clustering can use HAProxy or LVS for client‑side balancing. ZeroMQ: decentralized; does not support load balancing. RocketMQ: supports load balancing; consumer count should not exceed queue count; similar to Kafka’s strategy. ActiveMQ: supports load balancing, can be implemented with ZooKeeper.

7. Clustering Model

Kafka: native leader‑slave stateless cluster; each broker can be both master and slave; relies on ZooKeeper for dynamic scaling. RabbitMQ: simple clustering with replication mode; advanced clustering less supported. ZeroMQ: decentralized; no clustering support. RocketMQ: typical master‑slave mode; name server is a lightweight stateless node; brokers register topics with name servers. ActiveMQ: simple master‑slave clustering; advanced clustering less supported.

8. Management UI

Kafka: average RabbitMQ: good ZeroMQ: none RocketMQ: none ActiveMQ: average

9. Availability

Kafka: very high (distributed) RabbitMQ: high (master‑slave) ZeroMQ: high RocketMQ: very high (distributed) ActiveMQ: high (master‑slave)

10. Message Duplication Guarantees

Kafka: at‑least‑once, at‑most‑once RabbitMQ: at‑least‑once, at‑most‑once ZeroMQ: no persistence; neither at‑least‑once nor at‑most‑once RocketMQ: at‑least‑once ActiveMQ: at‑least‑once

11. Throughput (TPS)

Kafka: very high (batch send/consume) RabbitMQ: relatively high ZeroMQ: very high RocketMQ: high (batch consume possible) ActiveMQ: relatively high

12. Subscription Model & Message Distribution

Kafka: topic‑based publish/subscribe with optional regex matching. RabbitMQ: four exchange types – direct, topic, headers, fanout. ZeroMQ: point‑to‑point (p2p). RocketMQ: topic/messageTag based publish/subscribe with regex matching. ActiveMQ: point‑to‑point and publish/subscribe modes.

13. Ordered Messages

Kafka: supported (order guaranteed per partition). RabbitMQ: not supported. ZeroMQ: not supported. RocketMQ: supported. ActiveMQ: not supported.

14. Message Acknowledgment

Kafka: supports producer acks (0, 1, all) and consumer offset commits. RabbitMQ: supports acknowledgments (autoAck true/false) and batch/async confirms. ZeroMQ: supported. RocketMQ: supported. ActiveMQ: supported.

15. Message Replay

Kafka: supports offset‑based replay. RabbitMQ: not supported. ZeroMQ: not supported. RocketMQ: supports time‑point replay. ActiveMQ: not supported.

16. Message Retry

Kafka: not built‑in but can be implemented via offset replay. RabbitMQ: not built‑in but can be achieved using manual acknowledgments. ZeroMQ: not supported. RocketMQ: built‑in retry with configurable intervals. ActiveMQ: not supported.

17. Concurrency

Kafka: high; each consumer runs in its own thread; consumer count limited by partition count. RabbitMQ: very high; Erlang‑based, can create multiple channels per connection. ZeroMQ: high. RocketMQ: high; similar consumer‑thread model as Kafka. ActiveMQ: high; can handle tens of thousands of messages per second per broker.