Predictive Approach for Database Schema Evolution

Information systems, including their core databases need to meet changing user requirements and adhere to evolving business strategies. Traditional database evolution techniques focus on reacting to change to smoothly perform schema evolution operations and to propagate corresponding updates to the data as effectively as possible. Adopting such posterior solutions to such changes generates high costs in human resources and financial support. We advocate an alternate solution: a predictive approach to database evolution. In this approach, we anticipate future changes during the standard requirements analysis phase of schema development. Our approach enables potential future requirements to be planned for, as well as the standard, determining what data is to be stored and what access is required. This preparation contributes significantly in the ability of the database schema to adapt to future changes and to estimate their relative costs.

Predictive approach for schema evolution

Hassina Bounif

In today's companies and organizations, databases are omnipresent. They are the means to represent and store information. Therefore, they must evolve whenever the information architecture (semantic and structure) changes. Databases are subject to evolution for several reasons that include the changes to the real-world or the emergence of new database user requirements. We may also consider non representational and non-functional aspects such as the technology and performance requirements. Those can also compel the database designer or administrator to change the database in order to be more optimal and adapted to those requirements. This thesis proposes to address the schema evolution problem by anticipating the potential changes for the evolution of the schema. The approach adopts the a priori perspective i.e. to plan in advance for a possible solution that makes a schema evolve over time. Such a solution works with both a set of assumptions and techniques. The elaborated approach is based on different components: • First, we propose to develop a schema repository which contains a wide number of heterogeneous schemas and their historical versions. This repository serves in a series of operations that are important for the study of the schemas such as selecting the pertinent schemas, building the change matrices and performing the data-mining process. • Second, we propose to develop a requirements ontology, meaning to say a domain ontology that describes the current concepts of a database as well as the concepts that represent the potential future requirements that could be part of the evolution. The method of construction of this ontology includes several phases such as the preparation of the data for this ontology, the establishment of the data-dictionary besides change perspectives that store the concepts that are associated to other concepts by relationships resulting from the application of the data-mining process, the evaluation of the domain ontology using a graph structure and finally the construction of the ontology using a set of procedures. • Third, we propose to develop a predicted schema which is a database schema that contains two types of metadata. The metadata, that represents current entities of the database and the metadata that represents potential future entities. The predicted schema has two methods of representation: the multi-representation strategy and the predicted design repository schemas strategy. The schemas with the multi-representation strategy are particular versions of schemas whereas the schemas with the predicted design repository schemas are particular repositories of schemas. • Finally, we propose a case study and an analysis of the approach. The case study to better understand how this approach works and the analysis to show its feasibility, the positive as well as negative aspects.

EPFL2009

Semantic validation in spatio-temporal schema integration

Anastasiya Sotnykova

This thesis proposes to address the well-know database integration problem with a new method that combines functionality from database conceptual modeling techniques with functionality from logic-based reasoners. We elaborate on a hybrid - modeling+validation - integration approach for spatio-temporal information integration on the schema level. The modeling part of our methodology is supported by the spatio-temporal conceptual model MADS, whereas the validation part of the integration process is delegated to the description logics validation services. We therefore adhere to the principle that, rather than extending either formalism to try to cover all desirable functionality, a hybrid system, where the database component and the logic component would cooperate, each one performing the tasks for which it is best suited, is a viable solution for semantically rich information management. First, we develop a MADS-based flexible integration approach where the integrated schema designer has several viable ways to construct a final integrated schema. For different related schema elements we provide the designer with four general policies and with a set of structural solutions or structural patterns within each policy. To always guarantee an integrated solution, we provide for a preservation policy with multi-representation structural pattern. To state the inter-schema mappings, we elaborate on a correspondence language with explicit spatial and temporal operators. Thus, our correspondence language has three facets: structural, spatial, and temporal, allowing to relate the thematic representation as well as the spatial and temporal features. With the inter-schema mappings, the designer can state correspondences between related populations, and define the conditions that rule the matching at the instance level. These matching rules can then be used in query rewriting procedures or to match the instances within the data integration process. We associate a set of putative structural patterns to each type of population correspondence, providing a designer with a patterns' selection for flexible integrated schema construction. Second, we enhance our integration method by employing validation services of the description logic formalism. It is not guaranteed that the designer can state all the inter-schema mappings manually, and that they are all correct. We add the validation phase to ensure validity and completeness of the inter-schema mappings set. Inter-schema mappings cannot be validated autonomously, i.e., they are validated against the data model and the schemas they link. Thus, to implement our validation approach, we translate the data model, the source schemas and the inter-schema mappings into a description logic formalism, preserving the spatial and temporal semantics of the MADS data model. Thus, our modeling approach in description logic insures that the model designer will correctly define spatial and temporal schema elements and inter-schema mappings. The added value of the complete translation (i.e., including the data model and the source schemas) is that we validate not only the inter-schema mappings, but also the compliance of the source schemas to the data model, and infer implicit relationships within them. As the result of the validation procedure, the schema designer obtains the complete and valid set of inter-schema mappings and a set of valid (flexible) schematic patterns to apply to construct an integrated schema that meets application requirements. To further our work, we model a framework in which a schema designer is able to follow our integration method and realize the schema integration task in an assisted way. We design two models, UML and SEAM models, of a system that provides for integration functionalities. The models describe a framework where several tools are employed together, each involved in the service it is best suited for. We define the functionalities and the cooperation between the composing elements of the framework and detail the logics of the integration process in an UML activity diagram and in a SEAM operation model.

EPFL2006

Pattern detection for conceptual schema recovery in data-intensive systems

Gianluigi Viscusi

In this paper, an approach for information systems reverse engineering is proposed and applied. The aim is to support a unified perspective to the reverse engineering process of both data and software. At the state of the art, indeed, many methods, techniques, and tools for software reverse engineering have been proposed to support program comprehension, software maintenance, and software evolution. Other approaches and tools have been proposed for data reverse engineering, with the aim, for example, to provide complete and up-to-date documentation of legacy databases. However, the two engineering communities often worked independently, and very few approaches addressed the reverse engineering of both data and software as information system's constituencies. Hence, a higher integration is needed to support a better co-evolution of databases and programs, in an environment often characterized by high availability of data and volatility of information flows. Accordingly, the approach we propose leverages the detection of object-relational mapping design patterns to build a conceptual schema of the software under analysis. Then, the conceptual schema is mapped to the domain model of the system, to support the design of the evolution of the information system itself. The approach is evaluated on two large-scale open-source enterprise applications. Copyright (c) 2014 John Wiley & Sons, Ltd.