All Products
Search
Document Center

AnalyticDB for MySQL:Elasticity of the computing layer

Last Updated:Dec 26, 2023

The computing layer of AnalyticDB for MySQL can be scaled on-demand or based on schedules.

In a storage/compute-decoupled architecture, AnalyticDB for MySQL makes compute nodes stateless by separating them from stateful storage nodes. This way, data is not migrated when compute resources are being scaled. The high elasticity of Alibaba Cloud's cloud infrastructures also helps AnalyticDB for MySQL to deliver a smooth scaling experience at the computing layer.

Scheduled scaling

AnalyticDB for MySQL can scale your cluster based on schedules, which can help you improve operational efficiency and reduce costs.

Without elasticity, you need to plan your resources based on business peaks, and large amounts of resources are idle during off-peak hours. AnalyticDB for MySQL Data Warehouse Edition (V3.0) provides the scheduled scaling feature that scales resources up at specified time points on specified days of the week to handle business peaks, and scales them back to reduce costs after the peaks end. AnalyticDB for MySQL even allows you to reduce the number of nodes in a resource group to zero during off-peak hours to help you reduce costs.

During scaling, AnalyticDB for MySQL Data Lakehouse Edition (V3.0) requests the desired resources from the inventory side. Then, the inventory side starts a scaling task and allocates the requested resources to the cluster.

The interactive resource group in Data Lakehouse Edition (V3.0) allows resources to match the actual business loads. The scheduled scaling feature of AnalyticDB for MySQL Data Lakehouse Edition (V3.0) delivers the following benefits:

  • Sufficient resource provision: The inventory for the scheduled scaling feature consists of a permanent resource group and an elastic resource group. The resources of the permanent resource group can be provisioned in 0.5 to 15 days. The inventory of this resource group is assessable. The resources of the elastic resource group can be provisioned in 4 to 180 seconds. They cost more, and the inventory cannot be assessed. The inventory side predicts the amounts of resources that will be required based on the amounts of used and available resources, and determines the amounts of resources to be added and released.

  • High scaling efficiency: The scheduled scaling feature prepares the required resources before the scaling window starts. This ensures a fast scaling speed.

  • Maximum resource utilization: You can configure multiple scaling windows for a cluster, and the system will automatically implement the scaling.

On-demand scaling

AnalyticDB for MySQL Data Lakehouse Edition (V3.0) provides the on-demand scaling feature by using the job resource group. In on-demand scaling mode, a cluster does not reserve any resources before queries are submitted. After a query is submitted, the query requests computing resources that are used exclusively for itself. You can also specify the amount of resources that are available for each query. This way, resources can be requested and billed on-demand, which decreases usage costs and enhances isolation between queries. This feature is suitable for scenarios that do not require fast response speed but involve large amounts of data. 2