Exploiting Design Information to Drive Object Distribution Models

Authors: Sandeep Purao, Hemant Jain, Derek Nazareth

Object-orientation and distributed systems are quickly becoming norms for new system development, generating renewed interest in distribution schemes traditionally directed at relational databases. Though a number of facilitating technologies, such as EJB, COM, or CORBA, are available, pragmatic, and rigorous approaches for effective distribution are still lacking. In this paper, we develop an approach for deriving object distribution models by exploiting design information. The approach exploits information realistically available during the design stage, without requiring guesstimates of operational parameters. It spans multiple layers, mappings across which may be adjusted depending upon changes in assumptions and operational conditions to create specific models. We demonstrate the approach by operationalizing it to create a model for object distribution in closely coupled client/server configurations. The detailed model we demonstrate, therefore, represents one specific instantiation of our approach. The model is accompanied by a decision support procedure that assists the designer in the search for a satisficing solution in the resulting combinatorial multiple criteria problem. We briefly describe a research prototype that served as proof-of-concept and demonstrate usefulness based on an analysis of results obtained for distribution of a real-world object-oriented application.