Majority of the global information systems are constructed from a number of heterogeneous distributed database systems that provide a global object-oriented view of the data stored at the remote systems. Such global information systems have two sides: the source side which consists of heterogeneous distributed databases and the global side which provides an integrated view of the database systems from the source side. User applications access data through the iterations over the classes of objects included in a global object-oriented view. The iterations over the classes of objects are implemented as iterations over the data items such as the rows in the relational tables on the source side of the system creating serious performance problems. This paper addresses the performance problem of object-oriented applications accessing data on the source side of global information system through an object-oriented view on a global side. We propose a number of transformation rules which allow for more efficient processing of object-oriented application on the source side. The rules can eliminate the iterations of classes of objects on the global schema side. We prove the correctness of the rules and show how to systematically apply the rule to object-oriented applications. The paper proposes a number of templates for programming of object-oriented applications that allow for easier and more efficient performance tuning transformations.