The same question, asked of every platform.
Read a row across: that is one operational question, and the term each broker answers it with. Read a column down: that is what an operator working on that platform actually deals with.
| Operational area | Apache KafkaFully supported | RedisFully supported | RabbitMQFully supported | Apache Artemis (formerly ActiveMQ Artemis)Fully supported |
|---|---|---|---|---|
| Core resourcesWhat you browse and operate on | Topics and partitions | Keys and streams | Exchanges and queues | Addresses, queues and subscriptions |
| Consumer modelHow work is shared out | Consumer groups | Stream consumer groups | Consumers | Consumers and durable subscriptions |
| Backlog signalThe number that says "behind" | Consumer lag | Pending entries / stream lag | Queue depth | Queue depth |
| Routing modelHow a message reaches a consumer | Topic and partition | Key / stream | Exchange and binding | Address and routing type |
| Message operationsWhat the console lets you do to a message | Browse and produce | Inspect and stream operations | Inspect and publish | Browse and publish |
| Access controlsWhat the broker itself enforces | ACLs | Redis permissions | Vhost and resource permissions | Broker address permissions |
Community covers Kafka and Kafka-compatible brokers — Redis, RabbitMQ, Apache Artemis and Memcached ship in Commercial. Where a broker has no counterpart for a concept, the honest answer is that the concept does not exist there — not that support is missing.
Everything above the broker is the same.
This is the half of the product that does not vary by platform. It is defined once and resolved the same way whichever broker a resource belongs to — which is what makes an access review one screen instead of one per broker (Access Review is a Commercial screen).
Three places the vocabularies refuse to line up.
These are the distinctions a unified console is most tempted to flatten. Each one changes what an operator should do next, so each one is kept.
Backlog is not one number
Kafka lag is a distance between offsets. A RabbitMQ or Artemis queue depth is a count of messages sitting in a queue. Redis has both a stream lag and a pending list, which answer different questions. Showing one "backlog" figure across all of them would be a lie that costs an incident.
Delivery is not one guarantee
A Kafka partition preserves order and lets a group re-read it. A RabbitMQ queue hands each message to one consumer and forgets it on acknowledgement. An Artemis durable subscription keeps copies for a subscriber that is not there. The console does not imply a shared guarantee none of them make.
Access control is enforced twice
BROKA's scoped role decides whether an action is offered. The broker's own ACL, permission or authorization decides whether it succeeds. Both layers stay visible, because a console that only showed its own model would tell you an operation was allowed right up until the broker refused it.

