Oracle R Connector for Hadoop has both client and BDA/Hadoop Cluster components.
On the client host, e.g., your laptop, there is an orhc package that is installed in a local R engine. It leverages client side Hadoop Cluster software to communicate with the Hadoop Cluster.
On the server side, e.g., Big Data Appliance or a non-Oracle Hadoop cluster, an R engine must be installed on each of the task nodes. There is an orhc-drv package that enables communication with the Hadoop cluster.
Oracle R Enterprise (which is licensed separately under the Oracle Advanced Analytics option to Oracle Database) client packages can optionally be installed on the cluster task nodes in R as well as the client host R engine to enable using Oracle R Enterprise features in conjunction with Oracle Database and/or Exadata.
From a ODI perspective there are 3 key components.
1. ODI Studio - This is the UI where you would design the solution for processing Big Data using the ODI declarative components and the Knowledge modules (connectors). There are knowledge modules that are specifically available for processing Big Data.
The ODI Studio is typically installed on your development machine - outside of your cluster.
2. ODI repository - This is where the ODI studio would connect to during development. The knowledge modules are imported into the repository prior to developing the solution and the solutions developed are persisted in the repository. The ODI repository is stored in a database (Oracle DB or MySql DB) which is typically installed on the master node in the cluster.
3. ODI agent - This is java process which is responsible for executing the solution or scenario (ODI's term for compiled version of the solution). Typically the ODI agent is installed on the master node in the cluster. The agent will connect to the repository and execute the solution or scenario.