All Products
Search
Document Center

ApsaraMQ for RabbitMQ:Instance types

Last Updated:Aug 26, 2026

ApsaraMQ for RabbitMQ offers five instance editions across two categories -- Serverless and Subscription -- each with a different deployment model, billing structure, and feature set. Choose an edition based on your isolation requirements, throughput needs, and budget.

Edition overview

Category

Edition

Deployment

Billing

Best for

Serverless

Shared

Shared cluster (logically isolated)

Pay-by-cumulative-usage

Variable workloads with unpredictable traffic

Serverless

Exclusive

Dedicated physical cluster

Reserved + Elastic

Steady workloads that need resource isolation

Subscription

Enterprise Edition

Shared cluster (logically isolated)

Subscription (upfront) with elastic TPS

Cost-effective production workloads on shared infrastructure

Subscription

Platinum Edition

Dedicated physical cluster

Subscription (upfront) with elastic TPS

Mission-critical workloads requiring maximum throughput and dedicated resources

Subscription

Professional Edition

Shared cluster (logically isolated)

Subscription (upfront) with elastic TPS

Lightweight workloads in supported regions

Important

New Professional Edition instances are available only in the China (Qingdao), China (Hohhot), and Japan (Tokyo) regions. In other regions, use Serverless instances with reserved and elastic specifications instead -- they provide better performance at lower cost. Existing Professional Edition instances continue to work and can be upgraded or downgraded.

Shared vs. dedicated deployment

  • Shared (logically isolated): Multiple tenants share the same physical cluster. Resources are logically isolated per tenant. If another tenant generates a traffic spike, your instance may experience transient performance fluctuations. Best when cost efficiency outweighs guaranteed performance isolation.

  • Dedicated (exclusive physical cluster): Your instance runs on dedicated hardware with no resource contention from other tenants. Network, compute, and storage resources are exclusively allocated to your workload. Best for workloads with strict latency, throughput, or compliance requirements.

Serverless Shared, Enterprise Edition, and Professional Edition use shared deployment. Serverless Exclusive and Platinum Edition use dedicated deployment.

Billing

Edition

Billing model

Details

Serverless Shared

Pay-by-cumulative-usage or Reserved + Elastic

With pay-by-cumulative-usage, you are charged based on the total number of messages sent and received. With Reserved + Elastic, you pay for reserved capacity, and usage beyond the reserved capacity is billed on a pay-as-you-go basis. See Billing of Serverless instances.

Serverless Exclusive

Reserved + Elastic

Pay for reserved capacity upfront. Usage beyond the reserved capacity is billed on a pay-as-you-go basis. See Billing of Serverless instances.

Enterprise / Platinum / Professional

Subscription

Pay upfront for a fixed term. Enable elastic TPS to increase peak throughput beyond the base specification. See Billing of subscription instances.

Availability and reliability

All editions provide the same high-availability and data-reliability guarantees:

  • Multi-zone high availability (HA): Instances are deployed across multiple availability zones.

  • Triplicate data persistence: All message data is stored with three replicas to prevent data loss.

  • SLA: 99.95% for most editions. Platinum Edition offers a 99.99% SLA.

Elastic capacity

Each edition supports elastic scaling beyond its base specification. The elastic headroom varies:

Edition

Elastic capacity limit

Serverless Shared

Up to 50,000 TPS

Serverless Exclusive

1x the base specification

Enterprise Edition

1x the base specification, up to 50,000 TPS

Platinum Edition

1x the base specification, up to 50,000 TPS

Professional Edition

0.5x the base specification

Feature comparison

Feature

Shared

Exclusive

Enterprise

Platinum

Professional

Dashboard (Prometheus + Grafana)

Yes

Yes

Yes

Yes

No

Global message routing

Yes

Yes

Yes

Yes

No

Encryption in transit (TLS v1.2), which lets clients verify the authenticity of the server

Yes

Yes

Yes

Yes

No

Encryption at rest

No

Yes

No

Yes

No

Version management (customize the version upgrade and release rules of an instance)

No

Yes

No

No

No

Notable differences:

  • Serverless Exclusive and Platinum Edition support encryption at rest. Only Serverless Exclusive supports version management.

  • Professional Edition does not support dashboards, global message routing, or encryption.

Resource limits

Cluster limits

Resource

Shared

Exclusive

Enterprise

Platinum

Professional

Vhosts

200

2,000

200

4,096

200

Exchanges

6,000

10,000

6,000

80,000

6,000

Queues

6,000

10,000

6,000

80,000

6,000

Bindings

20,000

20,000

20,000

No limit

20,000

Queues per exchange

256

256

256

256

256

Max message size

10 MB

50 MB

10 MB (1 MB for instances in China (Qingdao), Japan (Tokyo), and China (Hohhot), and for instances created before May 1, 2025 in other regions)

50 MB (3 MB for instances in China (Qingdao), Japan (Tokyo), and China (Hohhot), and for instances created before May 1, 2025 in other regions)

64 KB

Max message retention

3 days

3 days

3 days

3 days

3 days

Max delivery attempts

16

16

16

64

16

Note

Exclusive and Platinum Edition support custom queue-per-exchange limits. Exclusive Edition users can submit a ticket to request customization. Platinum Edition users should contact Alibaba Cloud technical support. After customization, Exclusive Edition instances cannot be downgraded to a shared cluster.

Connections

Edition

Connection limit

Shared

10,000

Exclusive

20,000

Enterprise Edition

Purchase by configuration. Free quota: 2,000. Maximum: 10,000.

Platinum Edition

Purchase by instance type. Free quota: 50,000. Maximum: 100,000.

Professional Edition

4x the number of queues. Range: 1,000 to 5,000.

Channels per connection

Edition

Channel limit per connection

Shared

2,000

Exclusive

3,000

Enterprise Edition

min(queue count, 2,048)

Platinum Edition

3,000

Professional Edition

max(queue count, 64)

Enterprise Edition: The channel limit equals the queue count, up to a maximum of 2,048. If you have fewer than 2,048 queues, the limit matches the queue count. If you have 2,048 or more queues, the limit is 2,048.

Professional Edition: The channel limit is at least 64. If you have more than 64 queues, the limit equals the queue count.

Consumers per queue

Edition

Consumer limit

Shared

Total consumers = queues x 4. Per queue: 800 to 5,000.

Exclusive

No limit

Enterprise Edition

Total consumers = queues x 4. Per queue: 800 to 5,000.

Platinum Edition

No limit

Professional Edition

Total consumers = queues x 4. Per queue: 50 to 5,000.

Delayed messages

Edition

Maximum delay

Shared

7 days

Exclusive

7 days (customizable via support ticket; after customization, the instance cannot be downgraded to a shared cluster)

Enterprise Edition

7 days

Platinum Edition

7 days (customizable via Alibaba Cloud technical support)

Professional Edition

24 hours

Consumption timeout

The consumption timeout defines how long a consumer can hold a message before acknowledging it. If a consumer exceeds this timeout, the channel closes and the unacknowledged message returns to the queue for redelivery.

Edition

Default

Maximum

Shared

5 minutes

3 hours

Exclusive

30 minutes

12 hours

Enterprise Edition

5 minutes

3 hours

Platinum Edition

30 minutes

12 hours

Professional Edition

1 minute

10 minutes

For more information, see Message timeout and retry.

API rate limits

The following rate limits apply to management operations. These limits do not affect message publishing or consuming throughput.

Operation

API

Shared / Enterprise / Professional

Exclusive / Platinum

Clear a queue

purgeQueue

500 TPS

No limit

Create an exchange

exchangeDeclare

500 TPS

No limit

Delete an exchange

exchangeDelete

500 TPS

No limit

Create a queue

queueDeclare

500 TPS

No limit

Delete a queue

queueDelete

500 TPS

No limit

Create a binding

queueBind

500 TPS

No limit

Delete a binding

queueUnbind

500 TPS

No limit

Recover messages

basicRecover

500 TPS

No limit

Requeue a message

basicReject(requeue=true), basicNack(requeue=true)

20 TPS

No limit

Choose an edition

If you need...

Choose

Why

Pay-per-use with no upfront commitment

Serverless Shared

Elastic capacity scales up to 50,000 TPS. No reserved resources to manage.

Dedicated hardware without a subscription

Serverless Exclusive

Reserved capacity on isolated infrastructure. Pay only for overages beyond your reservation.

Cost-effective subscription capacity on shared infrastructure

Enterprise Edition

Subscription pricing with elastic TPS. Runs on a shared cluster to reduce cost.

Maximum performance and highest SLA

Platinum Edition

Dedicated hardware. Supports up to 80,000 queues, 100,000 connections, and 64 delivery retries. 99.99% SLA.

Lightweight workloads in select regions

Professional Edition

Available only in China (Qingdao), China (Hohhot), and Japan (Tokyo). For new deployments in other regions, Serverless instances provide better performance and cost-effectiveness.

Domestic localization support

ApsaraMQ for RabbitMQ supports Hygon processors to meet China's domestic localization requirements.