We’re excited to announce that Oracle Database In-Memory is now available on Autonomous Database (ADB) Serverless.
Database In-Memory adds in-memory database functionality to existing Oracle databases, and transparently accelerates analytics by orders of magnitude while simultaneously speeding up mixed-workload OLTP. Database In-Memory can accelerate analytic processing for any existing Oracle database compatible application – no application changes are required.
ADB Serverless environments can automatically take advantage of Database In-Memory and by default can use up to 50% of the SGA to populate objects in the In-Memory (IM) column store. All that is required is to enable objects for in-memory. The IM column store will grow or shrink automatically to accommodate enabled objects based on workload. However, this will not happen instantaneously, it will require some time and a continuous query workload. The total size of the column store can be adjusted larger or smaller than the default value of 50% of the SGA. It can be adjusted from 0 to 70 percent of the SGA size using the procedure DBMS_INMEMORY_ADMIN.SET_SGA_PERCENTAGE.
The ADB Serverless environment also supports Automatic In-Memory (AIM). This means that if the IM column store does fill up to the maximum specified by the SET_SGA_PERCENTAGE procedure then AIM will manage the contents of the IM column store. It will use access tracking, column and other relevant statistics to determine which objects will get the most benefit from being populated in the IM column store.
It is possible that even with AIM enabled one or more objects may only be partially populated. If this occurs then the data that is populated will be accessed in the IM column store, and the rest of the data can be accessed from the Exadata Smart Flash Cache. The feature that allows this is called Cross-Tier Columnar Scans and essentially means that data from one object can be scanned from memory or flash, wherever the data has been populated.
For more information about how Database In-Memory works in ADB Serverless see the Using Oracle Autonomous Database Serverless documentation.
