When an interface is created, ODI will by default create the staging area on the target server. There are many reasons for this design choice:
Data is usually loaded into servers that are larger than the ones data is extracted from: data warehouses for instance will host a lot more data than individual systems, and as a result these systems will tend to offer a lot more processing power.
Centralizing data coming from disparate sources is more convenient in a central location. As long as everything has to be loaded on the target system eventually, staging on the target will save us additional data movement.
Often times, source systems cannot be used for staging because of restricted permissions (read only).
However, there are cases where staging data away from the target is necessary. For example, if a process writes to a flat file, there is no way to transform data by leveraging the engine on the target side: there is none to use...