Instance editions provided by ApsaraMQ for Kafka - ApsaraMQ for Kafka

This topic describes the instance editions of ApsaraMQ for Kafka to help you select a suitable instance edition.

Editions

Note

Single-zone deployment: The service and data are deployed in a single zone. If a zone-level failure occurs in this mode, the service will be unavailable, and data may be lost. If your ApsaraMQ for Kafka instance is deployed in this mode, we recommend that you create another instance in another region, then use the connector ecosystem integration feature of ApsaraMQ for Kafka to back up messages in your instance. For more information, see Best practice for single-zone disaster recovery.
Multi-zone deployment: The service and data are deployed in multiple zones. This mode can help prevent service unavailability and data lose caused by zone-level failures in single-zone deployment mode. We recommend that you use this mode for key business.

The following table describes the instance editions of ApsaraMQ for Kafka.

Item	Standard Edition (High Write)	Professional Edition (High Write)	Professional Edition (High Read)	Serverless Basic Edition	Serverless Standard Edition	Serverless Professional Edition
Storage costs	Storage space is purchased in reserved mode. The storage costs are almost the same as the storage costs of self-managed Apache Kafka clusters. If you purchase a disk of 300 GB, the actual storage space that you can use to store your business data is 100 GB. The remaining 200 GB is used to store backups.	Storage space is purchased in reserved mode. The storage costs are 66% less than the storage costs of self-managed Apache Kafka clusters. If you purchase a disk of 300 GB, the actual storage space that you can use to store your business data is 300 GB. Alibaba Cloud provides an additional 600 GB of storage space free of charge to allow you to store backups.		The pay-as-you-go billing method is used. You are charged based on the actual storage space and duration. The storage costs are at least 70% less than the storage costs of self-managed Apache Kafka clusters.
Computing costs	Computing power is purchased in reserved mode.	Computing power is purchased in reserved mode.		The pay-as-you-go billing method is used.
Service interruption	When a cluster is scaled out, the system can quickly handle unexpected traffic spikes without data replication.
Scalable architecture	After a scale-out is triggered, scalability in seconds is supported for new read and write operations.			The storage-compute separation architecture helps implement the scalability of message reads and writes and the migration of partitions in seconds.
Lossless/seamless vertical scaling	Not supported.	Not supported.	Not supported.	You can manually upgrade or downgrade instance configurations to implement elasticity.	Lossless elastic traffic that is twice the reserved capacity is supported.	If the reserved capacity of an instance is small, the maximum elastic traffic for message production and consumption is 1,024 MB/s. The maximum elastic traffic is calculated by using the following formula: Maximum elastic traffic = `Max (1,024, Reserved capacity × 2)`. Note `Max (a, b)` indicates that the larger value of a and b is used.
Ratio of maximum read traffic to maximum write traffic	1:1	1:1	3:1	n:1
Instance type	Virtual instances whose specific resources are shared.	Dedicated instances.		Virtual instances whose specific resources are shared.		Dedicated instances.
Topic TTL	Not supported.	Supported only by topics that use local storage.		Supported.
Message retention period	Up to 7 days.	Custom configuration is supported.		No limit is imposed on the storage duration of messages. By default, messages are stored on ApsaraMQ for Kafka brokers for one year. If you want to store messages for a longer period of time, submit a ticket.
Disaster recovery	Compute and storage nodes are deployed in single-zone mode.	Instances can be deployed in multi-zone mode. If you select the single-zone mode when deploying an instance, compute and storage nodes are deployed in single-zone mode.		Compute and storage nodes are deployed in single-zone mode.		Three-zone deployment.
Performance optimization	Not supported.	Custom configuration is supported.		Custom configuration is supported.
Access control list (ACL)	Not supported.	Supported.		Supported.
SSL-based data encryption in virtual private clouds (VPCs)	Not supported.	Supported.		Supported.
Cross-zone deployment	Not supported.	Supported.		Not supported.		Supported.
Resource isolation	Specific resources, such as disks, are shared. Resource sharing may cause jitters, which are indicated by high latency. For example, unexpected traffic spikes may cause high loads on disks and thus increased request latency and timeout errors. Clients support retries to prevent message loss. The retry feature does not affect messages that have been written.	Exclusive clusters.		Exclusive clusters.


Compatibility with client versions	Client versions 0.11 to 3.x are supported.	Client versions 0.11 to 3.x are supported.		Client versions 0.11 to 3.x are supported.
Service-level agreement (SLA)	99.95%	If you set the Cross-zone Deployment parameter to Yes, an SLA of 99.99% is provided. If you set the Cross-zone Deployment parameter to No, an SLA of 99.95% is provided.		An SLA of 99.9% is provided. Compared with serverless Standard Edition and Professional Edition instances, this edition of instances uses more low-cost resources, including HDDs, Object Storage Service (OSS) resources, and preemptible Elastic Compute Service (ECS) instances. We recommend that you use this edition of instances in tests or scenarios with stable traffic. If you require higher business stability, we recommend that you use serverless Standard Edition and Professional Edition instances.	An SLA of 99.95% is provided. We recommend that you use this edition of instances in the production environment.	An SLA of 99.99% is provided, three-zone disaster recovery is supported, and high elastic traffic is supported for instances with small reserved capacity. We recommend that enterprises use this edition of instances.