All Products
Search

This Product

    Realtime Compute for Apache Flink:Billable items

    Document Center

    Realtime Compute for Apache Flink:Billable items

    Last Updated:Aug 19, 2026

    Realtime Compute for Apache Flink charges for the compute units (CUs) your workspaces consume. This topic explains the metering unit, what gets billed, which billing methods are available, and what additional costs you may incur from related Alibaba Cloud services.

    Metering unit

    The billing unit is the compute unit (CU). One CU equals 1 CPU core, 4 GiB of memory, and 20 GB of local storage. Local storage is used for logs and checkpoints.

    The number of CUs a deployment consumes depends on the queries per second (QPS) of input data streams, the computing complexity, and the data distribution. Use the following throughput benchmarks to estimate your CU needs.

    Scenario

    Throughput per CU

    Simple streaming (filtering, cleansing)

    40,000–55,000 records/second

    Complex streaming (aggregations, complex user-defined function (UDF) calculations)

    5,000–10,000 records/second
    These benchmarks cover only Flink's internal processing. External data read/write performance is not included and may lower effective throughput.
    If Simple Log Service query quotas are limited, Flink throughput is capped by those limits.
    If ApsaraDB RDS has connection or transactions per second (TPS) limits, Flink throughput is capped by RDS throttling.
    Window functions are compute-intensive. For workloads that use window functions, purchase at least 4 CUs.

    Billable items

    Each workspace has two billable items.

    Billable item

    Description

    Management resources

    Each workspace includes a development console. The console and its components consume approximately 2 CUs for management overhead.

    Computing resources

    Resources consumed when running Flink jobs, metered in CUs. Charged under the billing method selected for the workspace.

    Billing methods

    Realtime Compute for Apache Flink supports three billing methods.

    Billing method

    Description

    Subscription

    Pay upfront for a fixed duration. The price is calculated based on the subscription period. For more information, see Subscription.

    Pay-as-you-go

    Pay for actual usage after the fact. Bills are generated hourly at the start of each hour (UTC+8). The new billing cycle starts after the bills are settled. For more information, see Pay-as-you-go.

    Hybrid billing

    Charges the sum of fixed resource fees and elastic resource fees. For more information, see Hybrid billing.
    Important

    For business-critical workloads, use subscription or pre-allocate resources. Pay-as-you-go jobs require available physical resources at start or restart time. If capacity in a region or zone is insufficient, jobs may fail to start, which affects continuity.

    Billing rules for related services

    Activating a Realtime Compute for Apache Flink workspace incurs only Flink service fees. When you activate and use Realtime Compute for Apache Flink, you may also be charged for other cloud services. The following table describes these services.

    Alibaba Cloud service

    When it applies

    Billing reference

    VPC

    A virtual private cloud (VPC) in the same region is required for every workspace.

    Billing

    OSS

    If you set the storage type to OSS Bucket when activating a workspace, deployment artifacts — checkpoints, savepoints, logs, and JAR packages — are stored in the associated bucket.

    Billing overview

    ARMS

    If you select Managed Service for Prometheus when purchasing a workspace, Application Real-Time Monitoring Service (ARMS) is automatically activated to provide monitoring and alerting.

    See below

    CloudMonitor

    If you select the free monitoring option when purchasing a workspace, CloudMonitor provides basic monitoring and alerting.

    If you select the free monitoring service when you activate a Flink workspace, CloudMonitor provides basic monitoring and alerting capabilities. For more information, see Free of charge.

    Free quotas

    ARMS billing details

    Monitoring fees:

    Important

    • The Managed Service for Prometheus instance is released when you release the Realtime Compute for Apache Flink workspace.

    • To stop using Managed Service for Prometheus without releasing the workspace, log on to the ARMS console and release the instance manually. For details, see Discard or restore metrics. After the instance is released, you can no longer use ARMS metrics for troubleshooting or configure monitoring and alerting in the Flink development console.

    • ARMS collects deployment metrics every 30 seconds. A higher collection frequency improves monitoring accuracy but increases cost.

    CloudMonitor notes

    After a workspace is released, a namespace is deleted, or a Realtime Compute for Apache Flink deployment is deleted, the associated metrics are also deleted.
    For a comparison of monitoring and alerting capabilities between CloudMonitor and ARMS, see Comparison between CloudMonitor and the Alert Management sub-service of ARMS.

    References