Research activity in 1997
Our work in 1977 was devoted to the development of methodologies and models for the design of information systems, starting from a deep analysis of the functional organisation of the enterprise in terms of a hierarchically structured set of organisation networks which describe the flow of objects (information or materials) manipulated in the enterprise, and its functional behaviour, i.e., the flow of control.
The first result in this field was M*, an information system design methodology developed for CIM environments which covers the analysis of a manufacturing organization (Organization Analysis), the modelling of its information system (Conceptual Design), and the design of the related database environments which will support the automated operations of the organization (Implementation Design). The model used for conceptual design was based on the integration of the Entity-Relationship model with PrT nets to model static and dynamic properties.
M* has been the subject of extensive experimentations by several consulting companies, which suggested improvements of the original design. The new version of the the Process and Data Net (PDN) model, which is used in the a methodology called M*-OBJECT, differs from the previous version in the following ways:
• the representation of behavioural properties of the system under analysis and the distinction between dynamic and behavioural aspects. In our approach, behavioural aspects concern system components capabilities, i.e., the intrinsic behaviour of components (what the system can do), while dynamic aspects are related to strategies which control the execution of enterprise activities that use system components, i.e., the extrinsic behaviour of the applications which manipulate system components, taking into account the environment in which the system is embedded (what has to be done);
• the introduction of the object-oriented paradigm to describe static and behavioural properties.
The object-oriented approach has gained tremendous interest because of its modelling power and expressiveness, and its capabilities for model reusability, modularity, flexibility and extendibility. Moreover, it encourages analysts to identify and classify real-world system components. The PDN model is now based on the object-oriented modelling paradigm for the following reasons:
• it introduces a few but powerful concepts, such as data abstraction, modularity, and object identity, to group into a class objects sharing a common structure and having the same behaviour. Composite objects (i.e., objects made up of simple and/or composite objects) are easily modelled;
• it stresses the representation of relevant system components as a collection of interacting objects. This approach allows analysts to easily manage evolutions of objects during their life-cycle;
• it provides the inheritance mechanism to organize objects into hierarchies and to enforce reusability of both data structures and basic operation specifications.
Despite its advantages, the object-oriented approach suffers from several weaknesses, particularly in the field of information system design:
• it addresses only the detailed design and coding phases of the information system life-cycle;
• it is very difficult to express interactions between enterprise activities (such as concurrency, parallelism, or causality);
• it is inadequate to deal with the construction of large collections of database applications.
The Petri net model also suffers from some drawbacks. The most important one is its lack of abstraction mechanisms to represent real-world complex objects.
The PDN model overcomes the previous limitations by supplying the designer with the correct abstraction regarding the particular properties of the system to be described. More specifically:
• to describe static and behavioural properties, the object-oriented approach is used with a rich set of data abstraction mechanisms;
• to describe activity interactions at any level of granularity, the high-level Petri net approach is used for dynamic system specification.
Major advantages of the PDN approach are:
• the PDN model is embedded in a complete methodology (M*-OBJECT), which provides methods and tools for information system analysis and design as well as for development of database applications. In this way, it becomes an effective tool to support modelling efforts of system designers;
• using PDN abstraction mechanisms for different system properties, PDN specifications provide a clearer and easier to understand interface language between business users and designers;
• the modelling process results in detailed executable specifications, which can be used for rapid prototyping. Hence, users can verify the behaviour of the system before the effective implementation of database applications starts.
Other research interests of the group are directed towards:
• conceptual access to and conceptual visualization of large heterogeneous multimedia infobases, integrated with boolean information and/or database retrieval
• data clustering strategies obtained by decoupling data clustering imposed by logical schemata from actual physical clustering
• adaptive optimization algorithms.
