|Practical Considerations for Network Computing - Classic OO can work, but perform miserably in distributed computing. The issue has to do with the amount of network traffic ('round trips') generate by accessing objects.
For each property or method that is accessed, at least one round trip occurs - that is in addition to the round trip to instantiate the object! For that reason, distributed components, such as those for MTS and DCOM, are usually designed to pass data and call methods in the least number of round trips.
Generally, designing objects with traditional interfaces and bundling those objects into components with interfaces optimized to minimize round trips will address this issue. Within the component, there is a control procedure that manages the objects, coordinates execution, and handles the exposed interface.
In effect, the component functions as a sort of batch processor. Knowledge and experience are the key factors in creating high-performance distributed applications.