IIS - Interoperable Information Systems
Background - Research activities - Subprojects - Publications
A large number of tomorrow's distributed information systems will be composed of pre-existing (legacy) systems. Applications at the global enviroment may access data in different component systems. The component systems are inherently different: they can be file systems, database systems and so on; they can be supported by different vendors with different release versions. Heterogeneity and autonomy are the most challenging issues of such distributed information systems.
Heterogeneity may arise in every aspect of the component systems, such as data model, data management, administration, etc. Autonomy may include execution autonomy (i.e., the behavior of a component system must not be interfered by the behavior of the global environment) and design autonomy (i.e., the component system must not be re-designed to accommodate the design of the global environment). We have been focusing on the issues of (1) data model heterogeneity and (2) heterogeneity and autonomy of transaction management subsystems.
It is widely accepted that object-oriented data models are particularly suitable as common models in the global environment. To provide data model transparency, data in the component systems must be exported to the common data model of the global environment. Apart from the various issues concerning syntactic and semantic conflicts among the component systems, one foundamental issue is related to supporting reliability of object identities in the global environment. We proposed a framework for object identity management in object-oriented views of component systems based on a classification of reliability of identities of data items in component systems.
The behavior of transaction management subsystems of the component systems can be significantly different. They may or may not support the same consistence requirements. They can be open or closed depending on whether they provide published API. The design and execution autonomy further complicates the issues. Even if the component systems behave in the same manner, the global environment might not be able to guarantee the same consistency level. Our research in transaction management has been divided in two parts.
First, we have studied new consistency criteria that are suitable in the global environment. We have proposed a generic correctness criterion, called Partition Serializability (PSR), which allows for the specification of a number of non-serializable correctness criteria, where each criterion can be used as a global correctness criterion. An IIS administrator may choose a global correctness criterion based on the consistency requirements on the global level. In an extension of PSR, different users working in the same environment may even choose different correctness criteria.
Meanwhile, we are investigating the possibility for the different component transaction management subsystems to co-exist and the possible consistency level that can be guaranteed in the global environment. In our approach, transactional behavior of component systems is encapsulated and exported to the global environment. Encapsulated behavior allows certain degree of context-based polymorphism. The global level consistency can be derived from the exported behavior.
For more information of the above and other research activities in the IIS project, please contact the people responsible for the subprojects.