Overview of In-Memory Column Store with Oracle RAC

Each node in an Oracle RAC environment has its own In-Memory (IM) column store. Oracle recommends that you equally size the IM column stores on each Oracle RAC node. For any Oracle RAC node that does not require an IM column store, set the INMEMORY_SIZE parameter to 0.

It is possible to have completely different objects populated on every node, or to have larger objects distributed across all of the IM column stores in the cluster. It is also possible to have the same objects appear in the IM column store on every node (on engineered systems, only). The distribution of objects across the IM column stores in a cluster is controlled by two additional sub-clauses to the INMEMORY attribute; DISTRIBUTE and DUPLICATE.

In an Oracle RAC environment, an object that only has the INMEMORY attribute specified on it is automatically distributed across the IM column stores in the cluster. The DISTRIBUTE clause can be used to specify how an object is distributed across the cluster. By default, the type of partitioning used (if any) determines how the object is distributed. If the object is not partitioned it is distributed by rowid range. Alternatively, you can specify the DISTRIBUTE clause to over-ride the default behavior.

On an Engineered System, it is possible to duplicate or mirror objects populated in memory across the IM column store in the cluster. This provides the highest level of redundancy. The DUPLICATE clause is used to control how an object should be duplicated across the IM column stores in the cluster. If you specify just DUPLICATE, then one mirrored copy of the data is distributed across the IM column stores in the cluster. If you want to duplicate the entire object in each IM column store in the cluster, then specify DUPLICATE ALL.