A PolarDB serverless cluster automatically scales compute resources based on workload demand. Instead of choosing fixed node specifications, you configure a scaling range -- the minimum and maximum number of read-only nodes and the PolarDB Capacity Unit (PCU) range per node. PolarDB scales the primary and read-only nodes within this range automatically.
A PCU is the unit of measure for serverless scaling and billing. Each PCU is approximately equal to 1 CPU core and 2 GB of memory. Serverless clusters scale per second and bill per second based on actual PCU consumption.
If you have an existing database and want to use the serverless feature, use Data Transmission Service (DTS) to migrate data to a new serverless cluster.
Prerequisites
Before you begin, make sure that you have:
An Alibaba Cloud account with access to the PolarDB console. For details, see Register and log on to an Alibaba Cloud account
(Optional) A Virtual Private Cloud (VPC) and vSwitch in the target region, if the cluster needs to connect to an existing Elastic Compute Service (ECS) instance over the internal network
Configuration decisions
Review the key configuration decisions before you start:
| Decision | Options | Default |
|---|---|---|
| Region and zone | Any supported region. Cannot be changed after creation | -- |
| Database engine | PolarDB for PostgreSQL 14, 15, or 16 | -- |
| Database edition | Enterprise Edition or Standard Edition | Enterprise Edition |
| Scaling range | Min/max read-only nodes (0--15), min/max PCUs per node (1--32) | -- |
| Storage type | Enterprise Edition: PSL5 or PSL4. Standard Edition: ESSD PL0, ESSD PL1, ESSD PL2, ESSD PL3, or ESSD AutoPL | Enterprise Edition: PSL5. Standard Edition: -- |
| High-availability Mode | Enterprise Edition: Enable, Off, or Three Zones. Standard Edition: Off only | Enterprise Edition: Enable. Standard Edition: Off |
| No-activity Suspension | On or Off. Pauses the cluster when idle. Only available for PostgreSQL 14 with Enterprise Edition | Off |
| VPC and vSwitch | Select an existing VPC or use the default | Default VPC |
Step 1: Configure basic settings
Go to the PolarDB cluster buy page.
Set Billing Method to Serverless.
Select a Region. Select a region close to your users or workloads to minimize network latency. The region cannot be changed after the cluster is created. For more information, see Regions and zones. > Important: Deploy the PolarDB cluster in the same region as the ECS instance it connects to. Cross-region connections go through the public Internet and degrade performance. Deploying in the same zone further reduces latency.
Set Creation Method to Create Primary Cluster.
Select the database engine. PolarDB for PostgreSQL 14, 15, and 16 are supported for serverless clusters.
Select the database edition. Enterprise Edition and Standard Edition are available. Enterprise Edition is selected by default.
Select the primary zone. A zone is an independent physical location within a region. All zones in a region provide the same service performance. You only need to select the primary zone -- the system automatically selects a secondary zone.
Configure the network. The network type is fixed to VPC. For optimal performance, place the PolarDB cluster in the same VPC as the ECS instance it connects to.
Use an existing VPC if you already have an ECS instance and the VPC meets your network requirements.
Use the default VPC and vSwitch otherwise: | Resource | Details | |----------|---------| | Default VPC | One per region. CIDR block uses a 16-bit subnet mask (for example, 192.168.0.0/16) with up to 65,536 private IP addresses. Does not count against your VPC quota | | Default vSwitch | One per zone. CIDR block uses a 20-bit subnet mask (for example, 192.168.0.0/20) with up to 4,096 private IP addresses. Does not count against your vSwitch quota |
Create a custom VPC if the defaults do not meet your requirements. See Create and manage a VPC.
Select the High-availability Mode. The available options depend on the database edition. Enterprise Edition offers Enable (default), Off, and Three Zones. Standard Edition supports Off only. When set to Off, only the primary cluster is deployed with three replicas of data. The storage unit price is half of what it would be with High-availability Mode enabled, but the Service-Level Agreement (SLA) is lower.
Set the scaling range.
Parameter Description Valid values Minimum Read-only Nodes Minimum number of read-only nodes. Set to at least 1 for high availability 0--15 Maximum Read-only Nodes Maximum number of read-only nodes. Scales automatically based on workload 0--15 Minimum PCUs per Node Minimum PCUs allocated to each node 1--31 Maximum PCUs per Node Maximum PCUs allocated to each node 1--32 Confirm PolarProxy Type. This is fixed to Standard Enterprise.
Select whether to enable No-activity Suspension. This feature is disabled by default. When enabled, PolarDB pauses the cluster after a period of inactivity with no database connections, reducing costs during idle periods. The cluster resumes automatically when a new connection arrives. > Note: No-activity Suspension is only available for PostgreSQL 14 with Enterprise Edition. This option does not appear for other engine versions or for Standard Edition.
Set Storage Type. The available storage types depend on the database edition you selected. Enterprise Edition: Standard Edition: ESSDs are ultra-high performance cloud disks built on a next-generation distributed block storage architecture. They support 25 Gigabit Ethernet and Remote Direct Memory Access (RDMA), delivering up to 1 million random read/write IOPS with low latency. For more details, see ESSDs. > Important: When ESSD storage is exhausted, the disk is locked and handles only read operations. If you select ESSD AutoPL, configure the Provisioned IOPS of AutoPL Disks parameter to boost IOPS beyond the initial maximum of 50,000. The provisioned IOPS parameter accepts a maximum value of 50,000, bringing the theoretical total to 100,000 IOPS.
Storage type Performance ESSD PL0 Baseline ESSD performance ESSD PL1 5x the IOPS and approximately 2x the throughput of ESSD PL0 ESSD PL2 Approximately 2x the IOPS and throughput of ESSD PL1 ESSD PL3 Up to 10x the IOPS and 5x the throughput of ESSD PL2. Best for high-concurrency, low-latency workloads ESSD AutoPL Decouples IOPS from storage capacity for flexible, cost-effective configuration Set Standard Edition Storage Capacity.
Step 2: Configure advanced options
Set the cluster name. Select Auto-generated to let the system assign a name after creation, or select Custom to specify your own. Custom names must meet these requirements:
2 to 256 characters in length
Starts with a letter
Can contain letters, digits, periods (.), underscores (_), and hyphens (-)
Cannot start with
http://orhttps://
Set Resource Group. Select a resource group from the drop-down list. A resource group is a logical container for organizing and managing Alibaba Cloud resources under your account. Each resource belongs to exactly one resource group. For more information, see Classify resources into resource groups and grant permissions on the resource groups.
Step 3: Confirm and purchase
Review all configurations.
Set Quantity. Up to 50 clusters can be created at a time.
Review the fee details and Terms of Service at the bottom of the page, then click Buy Now to complete the purchase.
Verify cluster creation
The cluster takes 10 to 15 minutes to provision after payment is completed. To verify:
Open the Clusters page in the PolarDB console. > Note: Select the correct region where your cluster is deployed. The cluster is not displayed if a different region is selected.
Check the cluster status. The cluster is available when the status changes from Creating to Running.
Alibaba Cloud provides a 50% launch discount for PolarDB for PostgreSQL serverless clusters. For more information, see Launch of the PolarDB for PostgreSQL serverless cluster.
Next steps
After the cluster is running:
Connect to the cluster: Set up a database connection from your application or client tool.
Monitor cluster performance: Use the Performance Monitoring feature in the PolarDB console to observe cluster behavior.
Review scaling settings: Adjust the PCU range and read-only node count as you observe actual workload patterns.