Oracle Cloud Infrastructure Documentation

Autonomous AI Database Serverless Billing for Database Tools

Shows billing information for Autonomous AI Database database tools for ECPU and OCPU compute models.

ECPU Compute Model

The following built-in tools have separate, configurable compute resource limits:

  • Spatial Studio

  • Graph Studio

  • Oracle Machine Learning

  • Oracle Data Transforms

  • Data Lake Accelerator

For these tools compute resource usage does not use the database’s base compute, as specified with the ECPU count, or any auto scaled ECPUs. These tools have separate ECPU allocation, independent from the database compute ECPU count (and independent of Compute auto scaling).

For Spatial Studio, Graph Studio, Oracle Machine Learning, and Oracle Data Transforms, the ECPU count and Maximum idle time configuration options let you specify compute resource allocation for the VMs that run the associated built-in tool. For Graph Studio and Oracle Machine Learning, the Estimated concurrent sessions configuration option lets you estimate the number of notebook sessions that can run concurrently for the tool. When you configure these built-in tools, the values for these configuration options mean the following in terms of resource usage and billing:

  • You do not pay for a tool's ECPU allocation if you do not use the tool.

  • The VMs associated with a tool are provisioned when you start using a tool. For example, if Graph Studio is disabled, billing does not begin when you enable the tool. Billing begins when you start using Graph Studio.

  • The ECPU count specifies the compute resource allocation that is dedicated to the tool. The value you enter for ECPU count applies in addition to the database ECPU count you specify for the instance.

    After you start using a tool you are billed per ECPU hour reserved from the time the built-in Tool is launched. ECPU usage is measured each second, in units of whole ECPUs and averaged across an hour. If your database is running for less than an hour or you change the number of ECPUs for only part of an hour, your built-in Tool ECPU usage is billed per second for the average ECPU consumption during that hour. The minimum ECPU consumption is one minute.

  • For Graph Studio and Oracle Machine Learning, the Estimated concurrent sessions field specifies the estimated number of notebook sessions that can run concurrently for a tool. Autonomous AI Database Serverless divides the tool’s ECPU allocation by estimated concurrent sessions value to calculate the per-session billing quantity for a notebook session.

  • If the calculated per-session billing quantity is less than 8 ECPUs, you are billed for 8 ECPUs per hour for the notebook session. If the calculated value includes a fraction, the value is rounded up to the next whole ECPU.

  • When multiple notebook sessions run concurrently, each active session is billed using the calculated per-session billing quantity. The total billed ECPU usage for the tool does not exceed the configured ECPU allocation for the tool.

  • Changes to Estimated concurrent sessions take effect immediately for billing. Estimated concurrent sessions applies only to Graph Studio and Oracle Machine Learning. It does not apply to Spatial Studio, Oracle Data Transforms, or Data Lake Accelerator.

  • Billing stops for a built-in tool's allocated ECPUs when the built-in tool is disabled, the instance stops or is terminated, or when the built-in tool is idle for more than the specified Maximum idle time. The default maximum idle times depend on the tool:

For example, assume you allocate 4 ECPUs as the base ECPU count for the Autonomous AI Database instance, with compute auto scaling disabled, and you allocate 100 ECPUs for Graph Studio with the maximum idle time set to 30 minutes and estimated concurrent sessions set to 10 (i.e. each Graph Studio notebook session will be billed 10 ECPUs for each billing hour). If 5 notebook sessions run for 30 minutes starting at 2 pm, your bill for the specified time period (2-3 pm) is a total of 54 ECPUs. This sum includes the 4 base ECPUs plus the additional tools usage of 50 ECPUs for Graph Studio (including 30 minutes of actual usage and 30 minutes of idle time).

As another example, assume you allocate 6 ECPUs as the base ECPU count for the Autonomous AI Database instance, with compute auto scaling disabled, and you allocate 96 ECPUs for OML and set estimated concurrent sessions to 10 with the maximum idle time set to 15 minutes. In this example, each OML notebook session will get 10 ECPUs (96 divided by 10 is rounded up to 10). Assume 10 notebook sessions run concurrently for 15 min starting at 2 pm and the database remains open for the hour between 2 to 3 pm. For the billing hour of 2-3 pm, you are charged 96 ECPUs for OML usage for 30 min (15 min actual usage plus 15 min idle time) and 6 ECPUs for the database compute resources for the full hour, for a total of 54 ECPUs assuming no other tools are used in this period.

For Graph Studio notebook session billing, assume you allocate 100 ECPUs for Graph Studio and set Estimated concurrent sessions to 10. Each Graph Studio notebook session is billed for 10 ECPUs for each billing hour. If 5 notebook sessions run concurrently during the billing hour, Graph Studio usage is billed for 50 ECPUs for that hour.

For Oracle Machine Learning notebook session billing, assume you allocate 96 ECPUs for OML and set Estimated concurrent sessions to 10. The per-session billing quantity is 10 ECPUs, because 96 divided by 10 is rounded up to 10. If 10 notebook sessions run concurrently during the billing hour, OML usage is billed for 96 ECPUs for that hour. The total billed ECPU usage for the tool does not exceed the configured ECPU allocation.

Note

Note: After you start using a tool with a specified ECPU count you are billed per ECPU hour reserved from the time the built-in Tool is launched. For Graph Studio and Oracle Machine Learning notebook sessions, billing uses the calculated per-session billing quantity ECPU usage is measured each second, in units of whole ECPUs and averaged across an hour. If your database is running for less than an hour or you change the number of ECPUs for only part of an hour, it is billed per second for the average ECPU consumption during that hour. The minimum ECPU consumption is one minute.

Depending on the workload type of the Autonomous AI Database instance, built-in tool usage is charged to the following SKUs:

  • Oracle Autonomous AI Transaction Processing – ECPU – BYOL

  • Oracle Autonomous AI Lakehouse – ECPU – BYOL

  • Oracle Autonomous AI JSON Database – ECPU

See ECPU Compute Model Billing Information for details on built-in tools billing.

OCPU Compute Model

Built-in tools do not have separate, configurable compute resource limits. The built-in tools use compute from the database’s selected OCPU count. Tool usage may also use auto scaled OCPUs, if necessary.