DocsICP FilingConsole
Documentation
Home Page PolarDB PolarDB for MySQL Billing Resource Plan Compute plans for serverless clusters

Compute plans for serverless clusters

Updated at:
Copy as MD
Product
My Favorites

PolarDB provides a compute plans for serverless clusters that offsets the cost of dynamic resources used by compute nodes for serverless clusters and the serverless feature of fixed-spec clusters, reducing your costs.

Notes

  • Purchasing or unsubscribing from a compute plans for serverless clusters only affects billing and does not interrupt your service.

  • A compute plans for serverless clusters is not region-specific and automatically covers the cost of dynamic resources for compute nodes in serverless clusters and for the serverless feature of fixed-specification clusters across all regions.

  • If the remaining capacity in your compute plans for serverless clusters is insufficient, any excess usage of dynamic compute node resources is charged at the pay-as-you-go rate.

  • You cannot renew, upgrade, or downgrade compute plans for serverless clusterss. If a compute plans for serverless clusters has insufficient capacity, is expiring, or has expired, you can purchase a new compute plans for serverless clusters, and its capacity is combined with that of any existing packages.

  • After you purchase a compute plans for serverless clusters, dynamic resource usage for compute nodes in serverless clusters and for the serverless feature of fixed-specification clusters is no longer billed hourly; instead, this usage is measured in PolarDB Capacity Units (PCUs) and deducted hourly from the compute plans for serverless clusters.

Pricing

Compute plans for serverless clusterss are priced uniformly at CNY 0.38 per PCU per hour in all regions, including the Chinese mainland, China (Hong Kong), and overseas regions. This is 95% of the standard serverless rate in the Chinese mainland. Deduction factors vary by region. See the deduction table below for details.

Compute package capacity

  • 30 to 5,000 PCU-hours

  • 5,000 to 100,000 PCU-hours

  • 100,000 to 1,000,000 PCU-hours

Deduction rules

  • PCU usage for all nodes in a cluster is deducted hourly.

  • For each node, the amount deducted (CU*H) per interval is calculated using the following formula: PCU count * (deduction factor * time factor). The time factor is the duration in seconds divided by 3600.

Note

  • Within an hour, a deduction is made from the compute plans for serverless clusters each time the PCU count of a node changes. If the PCU count of the node does not change within the hour, one deduction is made from the compute plans for serverless clusters based on the node's current PCU count.

  • For each node, the sum of its time factors over a one-hour period is 1. This ensures that the total hourly deduction is calculated using the full regional deduction factor.

image

Example 1: PCU count remains unchanged within one hour

Consider a cluster in the Chinese mainland, where the regional deduction factor is 1. The Serverless configuration is set with a minimum resources per node of 1 PCU and a minimum number of read-only nodes of 1. If the cluster has no active connections during a one-hour period, it will have one primary node (1 PCU) and one read-only node (1 PCU). According to the deduction rules, the total deducted usage for the cluster in that hour is calculated as follows: the primary node usage (1 (PCU count) * 1 (deduction factor) * 1 (time factor)) plus the read-only node usage (1 (PCU count) * 1 (deduction factor) * 1 (time factor)), totaling 2 CU*H.

Example 2: PCU count changes within one hour

Consider a cluster in China (Hong Kong), where the regional deduction factor is 1.9. The Serverless configuration is set with a minimum resources per node of 1 PCU, a maximum resources per node of 16 PCU, a minimum number of read-only nodes of 1, and a maximum number of read-only nodes of 4. If the cluster workload increases during an hour, the PCU count for its nodes changes as follows:

Node

PCU count

Duration (seconds)

Deduction (CU*H)

primary node

1

2700

1.425 = 1 (PCU count) * 1.9 (deduction factor) * 2700 / 3600 (time factor)

1.5

90

0.07125 = 1.5 (PCU count) * 1.9 (deduction factor) * 90 / 3600 (time factor)

2

90

0.095 = 2 (PCU count) * 1.9 (deduction factor) * 90 / 3600 (time factor)

2.5

90

0.11875 = 2.5 (PCU count) * 1.9 (deduction factor) * 90 / 3600 (time factor)

3

90

0.1425 = 3 (PCU count) * 1.9 (deduction factor) * 90 / 3600 (time factor)

3.5

540

0.9975 = 3.5 (PCU count) * 1.9 (deduction factor) * 540 / 3600 (time factor)

read-only node (no horizontal scaling)

1

2700

1.425 = 1 (PCU count) * 1.9 (deduction factor) * 2700 / 3600 (time factor)

1.5

180

0.1425 = 1.5 (PCU count) * 1.9 (deduction factor) * 180 / 3600 (time factor)

2

180

0.19 = 2 (PCU count) * 1.9 (deduction factor) * 180 / 3600 (time factor)

2.5

540

0.7125 = 2.5 (PCU count) * 1.9 (deduction factor) * 540 / 3600 (time factor)

Based on the table, the cluster's total deducted usage for this hour is the sum of the primary node's usage (1.425 + 0.07125 + 0.095 + 0.11875 + 0.1425 + 0.9975) and the read-only node's usage (1.425 + 0.1425 + 0.19 + 0.7125), totaling 5.32 CU*H.

Deduction table

Deduction factors by region

Region

Product version

Deduction factor

Deductible capacity

Chinese mainland

Enterprise

1

50 CU*H / 1 = 50 CU*H

Standard

0.875

50 CU*H / 0.875 = 57.14 CU*H

US (Virginia)

Enterprise

1.3

50 CU*H / 1.3 = 38.46 CU*H

Standard

1.1375

50 CU*H / 1.1375 = 43.96 CU*H

US (Silicon Valley)

Enterprise

1.55

50 CU*H / 1.55 = 32.26 CU*H

Standard

-

-

Japan (Tokyo)

Enterprise

1.6

50 CU*H / 1.6 = 31.25 CU*H

Standard

1.4

50 CU*H / 1.4 = 35.71 CU*H

Korea (Seoul)

Enterprise

1.6

50 CU*H / 1.6 = 31.25 CU*H

Standard

-

-

Germany (Frankfurt)

Enterprise

1.7

50 CU*H / 1.7 = 29.41 CU*H

Standard

1.4875

50 CU*H / 1.4875 = 33.61 CU*H

Indonesia (Jakarta)

Enterprise

1.75

50 CU*H / 1.75 = 28.57 CU*H

Standard

1.53125

50 CU*H / 1.53125 = 32.65 CU*H

Philippines (Manila)

Enterprise

1.75

50 CU*H / 1.75 = 28.57 CU*H

Standard

-

-

Thailand (Bangkok)

Enterprise

1.75

50 CU*H / 1.75 = 28.57 CU*H

Standard

-

-

Singapore

Enterprise

1.9

50 CU*H / 1.9 = 26.32 CU*H

Standard

1.6625

50 CU*H / 1.6625 = 30.08 CU*H

Malaysia (Kuala Lumpur)

Enterprise

1.9

50 CU*H / 1.9 = 26.32 CU*H

Standard

-

-

China (Hong Kong)

Enterprise

1.9

50 CU*H / 1.9 = 26.32 CU*H

Standard

1.6625

50 CU*H / 1.6625 = 30.08 CU*H

UK (London)

Enterprise

1.9

50 CU*H / 1.9 = 26.32 CU*H

Standard

-

-

Application order

If you purchase multiple compute plans for serverless clusterss, the compute plans for serverless clusterss are deducted in order of their expiration time from earliest to latest. If the expiration time is the same, the compute plans for serverless clusters that was purchased first is deducted first.

Best practice

This example shows how to estimate the required capacity for a compute plans for serverless clusters.

  • Example PolarDB requirements:

    • Region: A region in the Chinese mainland.

    • Clusters:

      • Clusters A, B, and C: One primary node (2 PCU) and one read-only node (2 PCU).

      • Cluster D: One primary node (4 PCU) and two read-only nodes (4 PCU each).

      • Cluster E: One primary node (2 PCU) and two read-only nodes (4 PCU each).

    Note

    In this example, the cluster workloads are stable, and node PCU counts do not change.

  • Estimated PCU consumption:

    Required capacity = (Total node PCUs) × deduction factor × number of hours

    • Clusters A, B, and C

      4 PCU (2 PCU + 2 PCU) × 1 (deduction factor) × 24 (hours) × 3 (clusters) = 288 PCU.

    • Cluster D

      12 PCU (4 PCU + 8 PCU) × 1 (deduction factor) × 24 (hours) = 288 PCU.

    • Cluster E

      10 PCU (2 PCU + 8 PCU) × 1 (deduction factor) × 24 (hours) = 240 PCU.

    In this example, the estimated daily consumption is 816 PCU (288 PCU + 288 PCU + 240 PCU), and the estimated monthly consumption is 24,480 PCU (816 PCU × 30). To account for potential workload fluctuations, consider adding a buffer of approximately 5%. Therefore, a compute plans for serverless clusters with a capacity of 25,704 PCU (24,480 PCU × 1.05) is sufficient for one month.

Related operations

Purchase compute plans for serverless clusters

Go to the PolarDB purchase page to purchase a compute plans for serverless clusters. Select the resource package capacity tier and resource package specifications, and then configure the quantity and purchase duration. You can confirm your order in the configuration summary panel on the right.

Note

Purchase duration is the validity period of the compute plans for serverless clusters. If you select one year, the compute plans for serverless clusters is valid for one year, after which any remaining capacity expires.

View compute plans for serverless clusters usage

Go to PolarDB console > Resource Package, and on the compute plans for serverless clusters tab, view your purchased compute plans for serverless clusterss and the Remaining Amount/Total Amount for each compute package.

Click Usage Details. You are redirected to the Cost Management > Account > Resource Package page to view the usage details of the current compute plans for serverless clusters.

Note

  • A compute plans for serverless clusters covers the hourly PCU usage of all nodes in a cluster. For more information about the deduction rules, see deduction rules.

  • A compute plans for serverless clusters takes effect immediately after purchase. However, deduction information may take up to six hours to appear. We recommend checking again later.

Unsubscribe from compute plans for serverless clusters

  1. You can go to the PolarDB console > Resource Packages page. On the Compute Plans for Serverless Clusters tab, find the target compute plans for serverless clusters, and click Unsubscribe in the Actions column.

  2. After you click Unsubscribe, you are redirected to the Cost Management > Subscription Orders > Resource Unsubscription page. Under the Unsubscribe Resources and Resource Packages options, find the target compute plans for serverless clusters and click Unsubscribe Resource.

Note

  • You cannot unsubscribe from a compute plans for serverless clusters if it has expired or its capacity has been exhausted.

  • For details on the refund process, see Refund Policy.

FAQ

Purchase and use

Covered fees

A compute plans for serverless clusters offsets the dynamic resource fees for compute nodes in serverless clusters and the Serverless feature of fixed-specification clusters.

Package usage

Note

If you have not purchased a serverless cluster or enabled the Serverless feature on your existing clusters, you can start using a compute plans for serverless clusters in two steps:

  1. Purchase a compute plans for serverless clusters.

  2. Create a serverless cluster or enable the Serverless feature for a fixed-specification cluster.

Required purchase

No. You can decide whether to purchase a compute plans for serverless clusters based on your needs. If you purchase one, it automatically offsets the compute resource fees of your serverless cluster.

Coupon usage

You can check the scope of your coupon to determine whether it applies to a compute plans for serverless clusters. If you confirm that the scope of your coupon includes the compute plans for serverless clusters, but you cannot apply it to the payment when you purchase a compute plans for serverless clusters, please contact us.

Total fee coverage

No. For a serverless cluster, a compute plans for serverless clusters only offsets compute node resource fees up to the package's capacity. data storage is billed separately, by default on a Pay-As-You-Go basis. You can also purchase a Storage Package to reduce storage costs.

Supported clusters

Supported clusters

Only Serverless clusters and fixed-specification clusters with the Serverless feature enabled can use the compute plans for serverless clusters. Other clusters do not support the compute plans for serverless clusters.

Can PolarDB clusters created before purchasing a compute plans for serverless clusters also use a compute plans for serverless clusters?

Yes.

If the Serverless feature was enabled on a fixed-specification cluster before a compute plans for serverless clusters was purchased, can the cluster still use a compute plans for serverless clusters?

Yes.

Sharing across clusters

Yes.

Applying to a specific cluster

Currently, you cannot apply a compute plans for serverless clusters to a specific cluster.

Covered usage

Pay-as-you-go cluster coverage

No. When you use a PolarDB cluster, you are charged for compute nodes, storage space, backup storage (for usage beyond the free tier), SQL Insight (optional), and cold data archiving (optional).

After you purchase a compute plans for serverless clusters, the compute nodes in your Serverless cluster are no longer billed by the hour. Instead, their usage is deducted hourly from the compute plans for serverless clusters based on PCU consumption. However, other fees cannot be deducted from the compute plans for serverless clusters.

Deduction calculation

  • PCU usage for all nodes in a cluster is deducted hourly.

  • For each node, the amount deducted (CU*H) per interval is calculated using the following formula: PCU count * (deduction factor * time factor). The time factor is the duration in seconds divided by 3600.

Note

  • Within an hour, a deduction is made from the compute plans for serverless clusters each time the PCU count of a node changes. If the PCU count of the node does not change within the hour, one deduction is made from the compute plans for serverless clusters based on the node's current PCU count.

  • For each node, the sum of its time factors over a one-hour period is 1. This ensures that the total hourly deduction is calculated using the full regional deduction factor.

Package duration estimate

The answer depends on your usage. Let's take a serverless cluster in the Chinese mainland as an example. The cluster has one primary node and two read-only nodes, each configured to scale between 1 and 32 PCUs.

  • Scenario 1: Suppose that from 09:00 to 19:00, the primary node runs at 4 PCUs, and the two read-only nodes each run at 2 PCUs. From 19:00 to 09:00 the next day, the primary node runs at 2 PCUs, and the two read-only nodes each run at 1 PCU.

    The daily usage is calculated as follows:

    • Primary node usage:

      • 09:00 to 19:00 (10 hours): 4 PCUs × 10 hours = 40 PCU-hours.

      • 19:00 to 09:00 (14 hours): 2 PCUs × 14 hours = 28 PCU-hours.

    • Read-only nodes usage (for two nodes):

      • 09:00 to 19:00 (10 hours): 2 PCUs/node × 2 nodes × 10 hours = 40 PCU-hours.

      • 19:00 to 09:00 (14 hours): 1 PCU/node × 2 nodes × 14 hours = 28 PCU-hours.

    The total daily deduction is 40 + 28 + 40 + 28 = 136 PCU-hours. Therefore, a 100,000 PCU-hour package would cover approximately 100,000 / 136 ≈ 735 days of Serverless compute node costs.

  • Scenario 2: Suppose that the primary node runs at 4 PCUs and the two read-only nodes each run at 2 PCUs. However, a 30-minute high-load period occurs from 12:00 to 12:30, during which the primary node scales up to 10 PCUs and each read-only node scales up to 8 PCUs.

    The daily usage is calculated as follows:

    • Primary node usage:

      • 12:00 to 13:00 hour: (10 PCUs × 0.5 hours) + (4 PCUs × 0.5 hours) = 7 PCU-hours.

      • Other 23 hours: 4 PCUs × 23 hours = 92 PCU-hours.

    • Read-only nodes usage (for two nodes):

      • 12:00 to 13:00 hour: [(8 PCUs × 0.5 hours) + (2 PCUs × 0.5 hours)] × 2 nodes = 10 PCU-hours.

      • Other 23 hours: (2 PCUs × 23 hours) × 2 nodes = 92 PCU-hours.

    The total daily deduction is 7 + 92 + 10 + 92 = 201 PCU-hours. Therefore, a 100,000 PCU-hour package would cover approximately 100,000 / 201 ≈ 497 days of Serverless compute node costs.

Deduction order

PolarDB clusters support three types of resource plans: fixed-specification compute plans, Serverless compute plans, and storage plans. You can purchase and stack multiple plans of each type.

  • Fixed-specification compute plans cover the fees for fixed-specification compute node resources of pay-as-you-go clusters in all regions under your account and are shared across these clusters.

  • A Serverless compute plan can be used to deduct the dynamic resource fees for compute nodes in Serverless clusters and the Serverless feature of fixed-specification clusters in all regions under your account (shared by multiple clusters).

  • Storage plans are available in two types: for Chinese mainland and for regions in China (Hong Kong) and outside the Chinese mainland. Each type of plan is shared by all PolarDB clusters within its specified regions.

If you have multiple plans of the same type, the following rules determine the deduction order and how excess usage is billed:

  • Plans are applied based on their expiration date; the one expiring soonest is used first. If multiple plans have the same expiration date, they are applied in the order they were purchased.

  • Once the capacity of all plans of the same type is exhausted, any remaining usage is automatically billed at the standard pay-as-you-go rate.

Upgrades and renewals

Upgrading the package

No. If your compute plans for serverless clusters has insufficient capacity, you can purchase another compute plans for serverless clusters. If you purchase multiple compute plans for serverless clusterss, they are deducted sequentially in the chronological order of the compute plans for serverless clusters purchases.

Renewing the package

No. Each compute plans for serverless clusters is valid for one, three, or five years. After the validity period ends, any remaining capacity in the compute plans for serverless clusters automatically expires. If a compute plans for serverless clusters is about to expire or has expired, you can purchase another compute plans for serverless clusters to use.

Other questions

Missing deduction details

A compute plans for serverless clusters takes effect immediately upon purchase, but deduction details may not appear for up to six hours. Please wait and check again later.

Unsubscribing from the package

You can unsubscribe from a compute plans for serverless clusters in the console. For detailed instructions, see Unsubscribe from a compute plans for serverless clusters.

Previous：Compute Plan for Clusters with Defined SpecificationsNext：Storage plans