Managed message queue services compared

A message queue service lets one part of a system hand work to another without waiting for it. Producers send messages, the service stores them, and consumers process them when they can. That decouples microservices and makes asynchronous processing practical: an order service publishes a message and moves on while a billing consumer works through the backlog. A queue also gives you retry and dead-letter handling for messages that keep failing, message ordering where a workflow needs it, persistence so a restart does not lose work, and a way to scale consumers up and down with load. A message broker adds routing, topics and protocols such as AMQP on top, and that is where the services below split.

These are cloud-native services with no broker to size or patch. You send and receive through the provider's API and pay for usage, which makes them the fastest route when producers and consumers live in the same cloud and the queue is a buffer, not a platform.

Amazon SQS offers standard queues (at-least-once delivery, best-effort ordering) and FIFO queues (ordered within a message group, exactly-once processing). A message can be up to 1 MiB and is kept from 1 minute to 14 days, 4 days by default. In US East the first million requests a month are free, then it is $0.40 per million standard requests and $0.50 per million FIFO requests, with each 64 KB chunk of a payload billed as one request (AWS pricing, quotas, AWS price list, published 11 September 2026).

Azure Service Bus is a fuller broker as a service: queues, topics and subscriptions over AMQP 1.0. Standard messages are limited to 256 KB; Premium runs on dedicated messaging units, accepts messages up to 100 MB over AMQP, and adds virtual network integration, geo-replication and JMS 2.0. East US list prices are $0.05 per million operations on Basic, a $10 monthly base on Standard, and $0.9275 per messaging unit-hour on Premium (quotas, Azure retail prices).

Google Pub/Sub bills throughput rather than requests: $40 per TiB after the first 10 GiB a month. Messages carry up to 10 MB of data, unacknowledged messages are retained for 7 days by default, and a topic can retain them for up to 31 days. Ordering keys preserve order per key, at up to 1 MBps of publisher throughput each (pricing, quotas).

The trade is portability. Each speaks its own API or, in Service Bus's case, AMQP 1.0, so an application written for RabbitMQ or Kafka clients does not move onto them unchanged. For a sharper SQS-versus-broker view, see RabbitMQ vs SQS.

A managed broker runs a real broker engine, so existing clients, exchanges, topics and consumer groups keep working. The provider runs the infrastructure in its own account, and the incidents that matter, a consumer that stops acknowledging, a queue that fills memory, a lagging consumer group, still land on your team.

Amazon MQ runs ActiveMQ Classic and RabbitMQ. For RabbitMQ it supports 4.3 and 4.2 on the mq.m7g instance type and 3.13 on mq.t3, mq.m5 and mq.m7g, and states that it does not support streams. In US East a single RabbitMQ mq.m7g.medium broker lists at $0.1367 an hour and a three-node mq.m7g.medium cluster at $0.41 an hour, plus $0.10 per GB-month of storage (versions, AWS price list, published 11 September 2026).

CloudAMQP hosts RabbitMQ and LavinMQ. The free RabbitMQ plan allows 100 queues, 20 connections and 1 million messages a month; the shared Tough Tiger plan is $19 a month; dedicated RabbitMQ starts at $50 a month for one node and $297 a month for three (CloudAMQP plans). The plans page describes CloudAMQP-hosted servers and does not list running nodes in a customer's own cloud account.

Confluent Cloud sells Apache Kafka by cluster type. Basic starts at $0 with the first eCKU free; Standard lists at $0.75 per eCKU-hour, with data in and out billed per GB and storage at $0.08 per GB-month. Confluent's own estimates assume about 70 percent utilization (Confluent pricing).

Amazon MSK is AWS-run Kafka: Standard brokers at $0.204 an hour for kafka.m7g.large, Express brokers at $0.408 an hour for express.m7g.large, and storage at $0.10 per GB-month in US East, or MSK Serverless priced per cluster-hour, partition-hour and GB (MSK pricing, AWS price list, published 14 October 2026).

For RabbitMQ providers head to head, including Amazon MQ and CloudAMQP, see RabbitMQ providers compared. For Kafka, see managed Kafka options.

Every option above, except AceMQ, runs the brokers in the provider's account, so messages leave your boundary. AceMQ's managed services start from the opposite premise. At Tier 3 AceMQ operates RabbitMQ, Kafka or Redis and Valkey inside your own AWS or Azure account or data centre: you keep the account, the cloud bill, the encryption keys and the data, and AceMQ takes the pager with response targets from 15 minutes, by plan; 24/7. At Tier 4 AceMQ runs them on its own platform for teams that would rather own no cloud account for messaging. Tiers 1 and 2 are 24/7 support, with proactive monitoring added at Tier 2, for teams that keep the pager.

It is priced by quote, because the cost depends on cluster count, commitments and whether the infrastructure is yours. A quote request is answered within 1 hour. AceMQ serves 160+ customers in 26 countries. It is not the cheapest route for a small, non-regulated workload that fits a shared plan or a pay-per-request queue, and this page says so.

If your team already runs the brokers well and wants a senior engineer behind it, you do not need a managed service at all. Tier 1 support answers tickets under a contracted SLA, and Tier 2 adds monitoring that warns you early while your team approves every action. It suits RabbitMQ, Kafka and Redis estates on virtual machines, Kubernetes or a cloud service, and it also covers the Redis-layer and broker-layer incidents that a cloud provider's support tier stops short of.

1. Is it a queue between services in one cloud, with no broker features needed? Use that cloud's queue: SQS, Service Bus or Pub/Sub.
2. Do your clients speak AMQP, JMS or the Kafka protocol, or do you need routing, replay or topics? Use a managed broker for the engine you run.
3. Must the messages stay in your own cloud account or data centre? That rules out the provider-account options, and leaves operating the brokers yourself or having a specialist operate them in your account.
4. Does nobody want the pager? Choose a managed service (Tier 3 or 4). If your team keeps it, choose support (Tier 1 or 2).

We compared eight managed options, AceMQ managed RabbitMQ, Amazon SQS, Amazon MQ, CloudAMQP, Azure Service Bus, Google Pub/Sub, Confluent Cloud and Amazon MSK, on six criteria: engine and protocol, where the brokers run, ordering and delivery, size and retention limits, list price shape, and when each is not a fit. Vendor facts were fetched on 26 October 2026 from each vendor's own documentation or public price list, and prices are US East list prices. AceMQ's are quoted per estate. AceMQ is one of the eight, so every option, AceMQ included, has a not-a-fit case. List prices exclude discounts and committed-use terms, so treat them as a starting point.