ApsaraMQ for Kafka offers six instance editions across two deployment models: reserved-capacity editions with fixed pricing, and Serverless editions billed by actual usage. Select an edition based on your workload type, availability requirements, and scaling needs.
Choose an edition
Use the following table to identify the edition that matches your workload.
| Edition | Deployment | Billing | SLA | Best for |
|---|---|---|---|---|
| Standard Edition (High-Write Version) | Shared (logically isolated) | Reserved | 99.95% | Cost-sensitive workloads with balanced read/write patterns |
| Professional Edition (High-Write Version) | Dedicated (exclusive physical cluster) | Reserved | Up to 99.99% | Production workloads that need multi-zone disaster recovery and high throughput |
| Professional Edition (High-Read Version) | Dedicated (exclusive physical cluster) | Reserved | Up to 99.99% | Read-heavy workloads with a 3:1 peak read/write ratio |
| Serverless Basic Edition | Shared (logically isolated) | Pay-as-you-go | 99.9% | Testing, development, or stable-traffic services |
| Serverless Standard Edition | Shared (logically isolated) | Pay-as-you-go | 99.95% | Production environments that need elastic scaling (recommended) |
| Serverless Professional Edition | Dedicated (exclusive physical cluster) | Pay-as-you-go | 99.99% | Enterprise services that need three-zone disaster recovery and high elasticity |
Reserved-capacity editions
Reserved-capacity editions require a throughput specification and storage size selected up front. You are billed by subscription or pay-as-you-go.
Standard Edition (High-Write Version) -- Shared instance with logical isolation. Single-zone deployment with a 1:1 peak read/write ratio.
Professional Edition (High-Write Version) -- Dedicated instance on an exclusive physical cluster. Supports multi-zone deployment, access control lists (ACLs), SSL encryption in a VPC, and customizable message retention. Peak read/write ratio: 1:1.
Professional Edition (High-Read Version) -- Dedicated instance optimized for read-heavy workloads with a 3:1 peak read/write ratio. Provides the same advanced features as the Professional Edition (High-Write Version).
Feature comparison
| Feature | Standard Edition (High-Write Version) | Professional Edition (High-Write Version) | Professional Edition (High-Read Version) |
|---|---|---|---|
| Deployment architecture | Shared instance (logically isolated) | Dedicated instance (exclusive physical cluster) | Dedicated instance (exclusive physical cluster) |
| Peak read/write ratio | 1:1 | 1:1 | 3:1 |
| Storage cost model | Reserved. Storage costs are comparable to a self-managed cluster. A 300 GB disk provides 100 GB for data and 200 GB for backups. | Reserved. Reduces costs by 66% compared to self-managed clusters. A 300 GB disk provides 300 GB for data plus 600 GB of free backup capacity. | Reserved. Reduces costs by 66% compared to self-managed clusters. A 300 GB disk provides 300 GB for data plus 600 GB of free backup capacity. |
| Service interruption during scale-out | When you scale out a cluster, the system can quickly handle traffic bursts without data replication. | When you scale out a cluster, the system can quickly handle traffic bursts without data replication. | When you scale out a cluster, the system can quickly handle traffic bursts without data replication. |
| Elastic architecture | After a scale-out, new read and write operations support elasticity in seconds. | After a scale-out, new read and write operations support elasticity in seconds. | After a scale-out, new read and write operations support elasticity in seconds. |
| Elastic scaling | Not supported | Not supported | Not supported |
| Disaster recovery | Single-zone deployment only | Multi-zone or single-zone deployment | Multi-zone or single-zone deployment |
| Message retention period | Up to 7 days | Customizable | Customizable |
| Topic TTL | Not supported | Supported for local storage | Supported for local storage |
| Performance tuning | Not supported | Customizable | Customizable |
| ACL | Not supported | Supported | Supported |
| SSL encryption in VPC | Not supported | Supported | Supported |
| Cross-zone deployment | Not supported | Supported | Supported |
| Disk encryption | Supported | Supported | Supported |
| Compatible client versions | 0.11 to 3.x | 0.11 to 3.x | 0.11 to 3.x |
| SLA | 99.95% | 99.99% (cross-zone) / 99.95% (single-zone) | 99.99% (cross-zone) / 99.95% (single-zone) |
Single-zone deployment keeps all compute and storage nodes in one zone. A zone-level failure can cause service unavailability and data loss. For production services, use multi-zone deployment. If you deploy a single-zone instance, create a backup instance in a different region and use the connector feature to replicate messages. For details, see Best practices for single-zone disaster recovery.
Serverless editions
Serverless editions separate storage from compute and bill based on actual usage. This architecture provides read and write elasticity in seconds and partition migration in seconds.
Serverless Basic Edition -- Shared instance with logical isolation. Uses cost-effective resources (HDDs, OSS, and Spot Instances), resulting in a 99.9% SLA. Best for testing or stable-traffic services.
Serverless Standard Edition (recommended) -- Lossless elastic throughput up to 2x the reserved capacity. Multi-zone deployment across three zones (3AZ). Recommended for production environments.
Serverless Professional Edition -- Dedicated instance on an exclusive physical cluster. Multi-zone deployment across three zones (3AZ). Instances with small reserved capacity have higher elasticity. The recommended enterprise-grade edition.
Feature comparison
| Feature | Serverless Basic Edition | Serverless Standard Edition | Serverless Professional Edition |
|---|---|---|---|
| Deployment architecture | Shared instance (logically isolated) | Shared instance (logically isolated) | Dedicated instance (exclusive physical cluster) |
| Peak read/write ratio | 3:1 | 3:1 | 3:1 |
| Storage cost model | Pay-as-you-go. Billed for the storage space and duration used. Saves more than 70% on storage costs compared to cloud disks for a self-managed cluster. | Pay-as-you-go. Billed for the storage space and duration used. Saves more than 70% on storage costs compared to cloud disks for a self-managed cluster. | Pay-as-you-go. Billed for the storage space and duration used. Saves more than 70% on storage costs compared to cloud disks for a self-managed cluster. |
| Compute cost model | Pay-as-you-go | Pay-as-you-go | Pay-as-you-go |
| Elastic architecture | Separation of storage and compute enables read and write elasticity in seconds and partition migration in seconds. | Separation of storage and compute enables read and write elasticity in seconds and partition migration in seconds. | Separation of storage and compute enables read and write elasticity in seconds and partition migration in seconds. |
| Elastic scaling | Manual upgrade or downgrade | Lossless elastic throughput up to 2x the reserved capacity | Elastic throughput calculated as Max(1024 MB/s, Reserved throughput x 2) |
| Disaster recovery | Single-zone deployment | Multi-zone deployment across three zones (3AZ) | Multi-zone deployment across three zones (3AZ) |
| Message retention period | Up to one year by default. Supports unlimited storage. To request a longer retention period, submit a ticket. | Up to one year by default. Supports unlimited storage. To request a longer retention period, submit a ticket. | Up to one year by default. Supports unlimited storage. To request a longer retention period, submit a ticket. |
| Topic TTL | Fully supported | Fully supported | Fully supported |
| Performance tuning | Customizable | Customizable | Customizable |
| ACL | Supported | Supported | Supported |
| SSL encryption in VPC | Supported | Supported | Supported |
| Cross-zone deployment | Not supported | Supported | Supported |
| Disk encryption | Not supported | Not supported | Not supported |
| Compatible client versions | 0.11 to 3.x | 0.11 to 3.x | 0.11 to 3.x |
| SLA | 99.9% | 99.95% | 99.99% |
The Max(a, b) function returns the greater of a and b. For example, if the reserved throughput for a Serverless Professional Edition instance is 400 MB/s, the elastic throughput limit is Max(1024 MB/s, 400 x 2 = 800 MB/s) = 1024 MB/s.
Partition limits
The number of partitions depends on the edition and throughput specification.
Serverless instances
All three Serverless editions share the same partition limits:
If the reserved production throughput is 1 GB/s or less, the cluster provides 3,000 partition replicas.
If the reserved production throughput exceeds 1 GB/s, 300 partition replicas are added for each additional 100 MB/s.
Maximum partition replicas per cluster: 30,000
Maximum partition replicas per topic: 600
To prevent instability from excessive message fragmentation, do not create too many partition replicas for a single topic. To request a higher limit, submit a ticket.
Standard Edition (High-Write Version)
| Throughput specification | Included partitions | Maximum partitions |
|---|---|---|
| alikafka.hw.2xlarge | 1,000 | 4,000 |
| alikafka.hw.3xlarge | 1,000 | 4,200 |
| alikafka.hw.6xlarge | 1,000 | 4,400 |
| alikafka.hw.9xlarge | 1,000 | 4,600 |
| alikafka.hw.12xlarge | 1,000 | 4,800 |
Professional Edition (High-Write Version)
| Throughput specification | Included partitions | Maximum partitions |
|---|---|---|
| alikafka.hw.2xlarge | 1,000 | 4,000 |
| alikafka.hw.3xlarge | 1,000 | 4,200 |
| alikafka.hw.6xlarge | 1,000 | 4,400 |
| alikafka.hw.9xlarge | 1,000 | 4,600 |
| alikafka.hw.12xlarge | 1,000 | 4,800 |
| alikafka.hw.16xlarge | 2,000 | 5,000 |
| alikafka.hw.20xlarge | 2,000 | 6,000 |
| alikafka.hw.25xlarge | 2,000 | 7,000 |
| alikafka.hw.30xlarge | 2,000 | 8,000 |
| alikafka.hw.60xlarge | 2,000 | 9,000 |
| alikafka.hw.80xlarge | 2,000 | 10,000 |
| alikafka.hw.100xlarge | 3,000 | 12,000 |
| alikafka.hw.120xlarge | 3,000 | 14,000 |
| alikafka.hw.150xlarge | 3,000 | 16,000 |
| alikafka.hw.180xlarge | 3,000 | 18,000 |
| alikafka.hw.200xlarge | 3,000 | 20,000 |
| alikafka.hw2.220xlarge | 4,000 | 24,000 |
| alikafka.hw2.300xlarge | 4,000 | 26,000 |
| alikafka.hw2.400xlarge | 4,000 | 28,000 |
| alikafka.hw2.500xlarge | 4,000 | 30,000 |
| alikafka.hw2.600xlarge | 5,000 | 32,000 |
| alikafka.hw2.700xlarge | 5,000 | 34,000 |
| alikafka.hw2.800xlarge | 5,000 | 36,000 |
| alikafka.hw2.900xlarge | 5,000 | 38,000 |
| alikafka.hw2.1000xlarge | 5,000 | 40,000 |
Professional Edition (High-Read Version)
| Throughput specification | Included partitions | Maximum partitions |
|---|---|---|
| alikafka.hr.2xlarge | 1,000 | 4,000 |
| alikafka.hr.3xlarge | 1,000 | 4,200 |
| alikafka.hr.6xlarge | 1,000 | 4,400 |
| alikafka.hr.9xlarge | 1,000 | 4,600 |
| alikafka.hr.12xlarge | 1,000 | 4,800 |
| alikafka.hr.16xlarge | 2,000 | 5,000 |
| alikafka.hr.20xlarge | 2,000 | 6,000 |
| alikafka.hr.25xlarge | 2,000 | 7,000 |
| alikafka.hr.30xlarge | 2,000 | 8,000 |
| alikafka.hr.60xlarge | 2,000 | 9,000 |
| alikafka.hr.80xlarge | 2,000 | 10,000 |
| alikafka.hr.100xlarge | 3,000 | 12,000 |
| alikafka.hr.120xlarge | 3,000 | 14,000 |
| alikafka.hr.150xlarge | 3,000 | 16,000 |
| alikafka.hr.180xlarge | 3,000 | 18,000 |
| alikafka.hr.200xlarge | 3,000 | 20,000 |
| alikafka.hr2.220xlarge | 4,000 | 24,000 |
| alikafka.hr2.300xlarge | 4,000 | 26,000 |
| alikafka.hr2.400xlarge | 4,000 | 28,000 |
| alikafka.hr2.500xlarge | 4,000 | 30,000 |
| alikafka.hr2.600xlarge | 5,000 | 32,000 |
| alikafka.hr2.700xlarge | 5,000 | 34,000 |
| alikafka.hr2.800xlarge | 5,000 | 36,000 |
| alikafka.hr2.900xlarge | 5,000 | 38,000 |
| alikafka.hr2.1000xlarge | 5,000 | 40,000 |