Oracle Autonomous Database is a self-managing database that delivers end-to-end automation of tasks traditionally performed by database administrations. It’s available in two infrastructure choices: Shared Exadata Infrastructure and Dedicated Exadata Infrastructure. With Shared, as the name implies, multiple tenants share Exadata machines. With Dedicated, customers have their Exadata isolated from other tenants. It offers more control on infrastructure, software, and maintenance lifecycle, including the separation of environments for production, development, testing, and so on.
A typical use case for Exadata deployment is database consolidation: The ability to overprovision CPU so that you can assign multiple databases to a physical core. This ability benefits running non-production databases, such as development or test environments, and other non-critical databases where the workload doesn’t justify a whole CPU core. Exadata allows companies to reduce cost without sacrificing the platform’s performance, availability, and security.
This blog describes how you can now achieve similar database consolidation on Autonomous Database Dedicated using partial OCPU instead of a whole OCPU. This feature applies to Autonomous Database on Dedicated Infrastructure and Autonomous Database on Cloud@Customer.
Fractional OCPU
Before Oracle Database Release Update 19.11, the maximum number of autonomous databases that you could provision was equal to the number of OCPUs on the Exadata Infrastructure, with the smallest database being a single OCPU. On an Exadata quarter rack with 100 OCPUs, 100 was the maximum number of Autonomous databases you could create. Another related constraint was the minimum storage size of 1 TB per database.
From Release Update 19.11 onwards, you can now provision autonomous databases using fractional OCPU and storage in GBs. With fractional OCPU, you can create databases with less than 1 OCPU using fractional units, from 0.1 to 0.9 OCPU, with up to 10 databases running on a single OCPU. Using the previous example, you can provision up to 1000 autonomous databases on an Exadata quarter rack. Along with fractional OCPUs, you can now provision storage in GB instead of TB, with a minimum storage of 32 GB and scaling increment of 1 GB. So, instead of assigning 1 TB for a 90 GB development database, Autonomous database users can now allocate 100 GB for better storage efficiency. Similar to autonomous databases with integer OCPU and storage allocation in TB, you can scale up or scale down Autonomous databases with fractional OCPU and storage allocation in GB with support for conversion from fractional to integer and GB to TB or the reverse with no downtime.
To create an Autonomous Transaction Processing database with fractional OCPU and storage allocation in GB, enter a decimal OCPU count between 0.1–0.9 and specify the storage size in GB on the Create Autonomous Database page.
All Autonomous Database features, such as autoscaling, cloning, and Autonomous Data Guard, are supported with fractional OCPUs. You can develop and test your production database functionalities such as disaster recovery failover with minimal OCPU and storage allocations.
As with integer OCPU Autonomous databases, fractional OCPU databases get performance-related resources allocated proportionally based on the number of OCPUs chosen. For example, an autonomous database with 0.3 OCPU gets a 30% allocation of the memory and concurrent sessions for a single OCPU Autonomous database.
A critical difference between fractional and integer OCPU is the availability of the predefined database services. These connection types are created at database creation time. Autonomous Data Warehouse with fractional OCPU supports only low service with its TCP and TCPS counterparts (low, low_tls, low_ro, low_ro_tls). Autonomous Transaction Processing with fractional OCPU supports both tp and low services.
Billing for fractional OCPU usage is similar to integer OCPU usage. The total active OCPUs (both integer and fractional) for an autonomous container database are aggregated and then rounded to the nearest integer OCPU. For example, a container database with 38 Autonomous Transaction Processing databases with 0.1 OCPU is billed as 4 OCPUs.
Want to know more?
Fractional OCPU and granular storage enable Autonomous Database users to reduce costs through database consolidation further. Utilizing the power of the Exadata Infrastructure allows greater consolidation density without sacrificing the platform’s performance, availability, and security. To get started, see the documentation.
Learn more about migrating to Autonomous Database and about reference architectures in Oracle Architecture Center.