Oracle Autonomous Database (ADB) recently introduced a more advanced ECPU billing metric and is now retiring the legacy OCPU billing metric for Autonomous Data Warehouse and Autonomous Transaction Processing in the next 12 months. Oracle recommend switching from the OCPU billing metric to the ECPU billing metric, which will not incur any downtime or service interruptions.
ECPUs will provide the same great price-performance as OCPUs and continuous improvements in price-performance over time. Updating to the ECPU billing metric provides the following benefits:
50% lower entry cost: The smallest Autonomous Database that can be provisioned with ECPUs is 50% less expensive ($0.672 per hour vs $1.3441 per hour with OCPUs)
Finer granularity for database scaling: Each incremental increase in ECPU database size is only $0.336
Lower storage costs: Autonomous Data Warehouse storage price reduced from $118.40 to $25.00 per TB per month and Autonomous Transaction Processing storage can be provisioned in increments of 1GB (this is a huge thing!), with a minimum of 20GB – this brings the ADW in-database storage price on par with the object storage and thus this helps to build data lakes solely on the architectural requirements and not focusing on cost
Up to 87% lower costs with database consolidation: Elastic Resource Pools, available on ECPU ADB Serverless databases, help consolidate deployments leading to major cost savings
New features for Autonomous Database may only be available with ECPU’s
Note that the prices mentioned above are the current list prices for Autonomous Databases with the License Included license type. Please, refer to the ECPU billing metric FAQ for more details about ECPUs and each of these benefits.
There are also differences in backups between OCPU’s and ECPU’s. ECPU’s backup storage is billed separately, and the backup retention period may be selected between 1 and 60 days. With OCPU’s, 60-days of backup storage is included in the storage price. This new ECPU customer-controllable backup is beneficial because customers can now control the backup storage size and further reduce the cost of dev/test environments. Here is how I reduced the size from 60 to 31 days (later on I did reduce it to 7 days).
I did scale down my database in 2 phases: (1) I switched to the ECPU model (1 OCPU –> 4 ECPUs) and then (2) reduced the ECPU count from 4 to 2 and the storage from 1024GB to 20GB (those two in one go with no downtime).
Here are some general guidelines related to the new ECPU metric:
Provision all new Autonomous Data Warehouse and Autonomous Transaction Processing databases or clones with the ECPU billing metric
Update all existing databases to the ECPU billing metric, which is a simple and seamless button click or API call
Note that if you choose not to update your existing databases’ billing metric at this time, Oracle may convert your databases from the OCPU billing metric to the ECPU billing metric in the future
Updating your Autonomous Database Serverless to the ECPU billing metric will have no impact to your service and incur no downtime. Oracle Autonomous Database will be retiring the OCPU-based SKUs and replacing them with the ECPU-based SKUs. Starting in August 2023, some new Autonomous Database features may be available only on ECPU’s. For example, Elastic Resource Pools are only available with ECPU’s.
Note that ECPU’s have also already been introduced for MySQL Heatwave on AWS, and other services may also offer ECPU’s in the future.
Here is some additional information regarding the ECPU announcement:
Autonomous JSON Database adopts ECPUs and per-gigabyte Storage, lowering the entry level price point for developers by Hermann Baer
Announcing The Retirement Of The OCPU Billing Metric In Autonomous Database Serverless: MOS Note Doc ID 2998742.1
For more information about the ECPU billing metric and its benefits refer to the ECPU FAQ
Oracle ECPU Documentation
Use and Manage Elastic Resource Pools on Autonomous Database
How to Achieve up to 87% Compute Cost Savings with Elastic Resource Pools on Autonomous Database
Read MoreJulian Dontcheff’s Database Blog
