No abstract

The recent boom of new application categories, such as multi-media systems, groupware, and the wide area distribution of information across the Internet, has led to further demands for flexibility in software. This paper presents a framework (COBS-OM) for building configurable parallel and distributed programs where type-dependent object functionality is explicitly separated from its characteristics subject to configuration. COBS-OM supports a programming model where dealing with configuration issues is a central part of the design. It provides abstractions for incorporating flexibility into a distributed object-oriented application in a methodical fashion. In addition, performance issues are addressed by considering runtime execution adjustments of the basic mechanisms that influence them. We introduce the basic elements of the model. We also present Data Object, a complex configurable object that provides flexible access to data output from a high performance parallel and distributed scientific application.