pax

The distribution of data within or between organizations is crucial for many applications. Such a distribution may be transparent, as it is embodied in distributed database systems. Its purpose is to increase the performance, reliability and availability of the system, but it also raisesits complexity. On the other hand, many applications either do not require transparent data distribution or even explicitly need their own copies of the data. Examples for applications and systems of this category are data warehouses which collect base data from different transaction processing systems, financial data distribution systems in banking where data are bought from external providers, stored in a data store under the control of the data owners and distributed from there to decentralized databases of different working groups, or applications in statistics that use derived data which are generated by individual researchers and made available to a broader audience by placing them into a database which can be accessed by different research groups. Replication systems are often used to accomplish the data distribution tasks as required for this kind of systems. Object-oriented concepts are becoming increasingly important to realize such applications. One often tends to store the data in object database systems. As a consequence, a replication system that is usable in the context of object databases is needed. However, today's replication systems do not sufficiently accommodate such an object-oriented environment. Even disregarding the incorporation of operations and concentrating on the structural aspects, an object replication system, compared to a relational replication system, has to be able to handle the additional complexity that is introduced by composite objects, inheritance and polymorphism. Applications as described above rely on the non-transparent replication of object databases. In addition, one does not replicate databases as a whole, but users can select which parts of a database are to be replicated. This thesis makes two main contributions. The first one is the development of a meta model that specifies the abstract functionality which is needed for the non-transparent, partial and selective replication of object databases. This meta model allows incorporation of its replication concepts in a variety of object models. The second contribution is the design of a replication infrastructure, based on the meta model, which serves as the foundation for concrete realizations of object replication systems. Before defining the meta model for our given problem domain, we need to generally define replicated objects and types. These definitions are based on a structural object model which is minimal; that is, the model can be considered the least common denominator of a variety of object models. We then summarize possible characteristics of replicated objects and types in a classification framework that serves as the basis for categorizing replication protocols, replication systems and replication-relevant application requirements. Next, an architecture with partially replicated databases is developed. It covers the needs of applications and systems as mentioned above and can thus be used to deduce concrete requirements that a suitable replication system has to meet. These results, namely, the definitions of replicated objects and types, the classification framework and the requirements for a replication system which is useful for the targeted application domain, lay the foundation forthe development of the replication meta model and the replication infrastructure that is based on this model. To demonstrate the validity and feasibility of the proposed solution, the implementation of a replication system for Calanda, an object-oriented time series management system, is presented. This working system realizes the main characteristics of the meta model and the replication infrastructure.

A meta model and an infrastructure for the non-transparent ...

A meta model and an infrastructure for the non-transparent replication of object databases

actions