2. The database kernel determines that the data is stored on Exadata cells so an iDB (Intelligent Database protocol, simply a unique Oracle data transfer protocol in between storage layer and DB layer ) command representing the SQL command is constructed and sent to the Exadata cells.
3. The Exadata Storage Server software scans the data blocks to extract only the relevant rows and columns that satisfy the SQL command.
4. Exadata cells then return to the database instance iDB messages containing the requested rows and columns of data not block images like traditional database storage (Ex: only 2MB out of 5GB).
5. The database kernel consolidates the result sets from across all the Exadata cells which is similar as parallel query consolidated operation.
6. The rows are returned to the client.