All Products
Search
Document Center

ApsaraDB RDS:Assured serverless

Last Updated:Mar 28, 2026

Assured serverless adds elastic scaling to your existing subscription or pay-as-you-go ApsaraDB RDS for PostgreSQL instances without changing their billing method. When traffic spikes, the instance automatically scales up CPU and memory to the configured upper limit. When the spike ends, resources scale back down to baseline — keeping costs in check without manual intervention.

How it works

Assured serverless monitors CPU and memory usage and responds to workload changes in two directions:

Scale-out (upgrade): When CPU or memory usage exceeds the configured upgrade threshold, the instance immediately scales up to the set upper limit. Unlike standard serverless, which steps up gradually, assured serverless jumps directly to the upper limit.

Scale-in (downgrade): After usage drops, the system enters a 90-second observation window. If both CPU and memory stay below the downgrade threshold for the full 90 seconds, the instance scales back down in steps — CPU decreases by 1 core per step, memory decreases by 2 GB per step. If either metric climbs back above the upgrade threshold during the window, the instance holds or scales out again.

    Note

    A transient connection interruption (up to 30 seconds) may occur if the host running your instance has insufficient resources and the instance must migrate to another host. To reduce this risk, set Effective Policy to Within Maintenance Window and schedule it during off-peak hours.

    Comparison with serverless instances

    Assured serverless is designed for teams that need elastic scaling on existing fixed-specification instances. Serverless instances (with the Serverless billing method) follow a different model.

    DimensionAssured serverlessServerless instances
    Billing methodSubscription or pay-as-you-go (unchanged)Serverless (RCU-based)
    Supported featuresPgBouncer connection pool, SSL encryption, Babelfish for RDS PostgreSQLSubset of features
    Scaling granularityCPU and memory adjusted independentlyRCU units only (fixed 1:2 CPU-to-memory ratio)
    Scale-out limitUp to 2x the original CPU and memoryMaximum 16 RCU
    Scale-up behaviorJumps directly to the upper limitSteps up gradually

      Prerequisites

      Before you begin, make sure:

      • The instance product type is Standard.

      • The billing method is Subscription or pay-as-you-go.

      • The instance is in a supported region and zone.

        Supported regions and zones

        RegionZone
        China (Hangzhou)Zone I, Zone J, Zone K
        China (Shanghai)Zone B, Zone E, Zone G, Zone L, Zone M, Zone N
        China (Shenzhen)Zone C, Zone E, Zone F
        China (Heyuan)Zone A, Zone B
        China (Guangzhou)Zone A, Zone B
        China (Beijing)Zone F, Zone G, Zone I, Zone J, Zone K, Zone L
        China (Zhangjiakou)Zone B, Zone C
        China (Ulanqab)Zone B, Zone C
        China (Chengdu)Zone A, Zone B
        China (Hong Kong)Zone B, Zone C, Zone D
        South Korea (Seoul)Zone A
        SingaporeZone A, Zone B, Zone C
        Malaysia (Kuala Lumpur)Zone A, Zone B
        Indonesia (Jakarta)Zone A, Zone B, Zone C
        Thailand (Bangkok)Zone A
        Germany (Frankfurt)Zone A
        UK (London)Zone A, Zone B

      Billing

      During the public preview, scaling up through assured serverless incurs no additional charges. You are billed only for the original instance specifications.

      Usage notes

      Enable assured serverless

      1. Go to the Instances page. In the top navigation bar, select the region where the instance resides. Find the instance and click its ID.

      2. On the Basic Information page, locate the Configuration Information section and click Settings next to Assured Serverless.

      3. On the Assured Serverless Settings page, turn on the feature switch and configure the parameters.

        ParameterDescription
        Resources threshold for assured serverlessSet the CPU Upper Limit and Memory Upper Limit for elastic scaling. Maximum CPU Resources: adjusts in increments of 1 core, up to twice the instance's initial CPU. Maximum Memory: adjusts in increments of 2 GB, up to twice the instance's initial memory.
        Effective policyChoose when the feature takes effect: Immediately, or Within Maintenance Window (the feature activates during the configured maintenance window). To view or update the maintenance window, see the Configuration Information section. If host resources may be constrained, select Within Maintenance Window and schedule during off-peak hours to reduce the chance of a transient connection interruption.
        Upgrade thresholdThe CPU Threshold and Memory Threshold that trigger scale-out. When either metric exceeds the threshold, the instance scales up directly to the upper limit.
        Downgrade thresholdThe CPU Threshold and Memory Threshold that trigger scale-in. After a 90-second observation window during which both CPU and memory remain below the thresholds, the instance scales down in steps. This dual-metric requirement prevents unnecessary scaling caused by brief single-metric spikes.
      4. Click OK.

      What's next