Policy Regulated Management Of Schema Evolution In Database-centric Environments

Abstract

Current information systems and specifically database systems are continuously evolving environments, where design constructs are added, removed or updated rather often due to changes to system specifications and additions to initial requirements. Database schema evolution is a very important and diachronic problem in the area of database research. It is a complicated issue, which is related to every phase of the development of an Information System. Several reasons during the development or operational phase of database system can trigger the modification of a populated database schema, such as schema changes accompanying changes of requirements, schema restructuring (i.e. normalization, de-normalization) due to performance reasons, redefinition of views, migration from a legacy system towards novel platforms, etc.Apart from the core database schema, database centric environments comprise a plethora of views and queries embedded in procedures, software modules, complex workflows, etc. that are also affected by evolution operations. Unfortunately, no support is provided from current RDBMSs for analyzing the impact and furthermore adjusting semantic and syntactical inconsistencies emerging on these parts, as results of such operations. Their reaction to evolution is still handled manually by administrators and developers. Although research has extensively dealt with the problem of database schema evolution investigating mainly the adaptation of internal database objects to schema changes, problems persist with existing queries and views, mainly due to the fact that in most cases, the proper attention is not given to their role as integral parts of the environment.In this thesis, we provide a representation technique that maps all essential constructs of a database centric environment to graphs. The basis of our framework is a graph model, called evolution graph, which models in a coherent and uniform way internal structural elements of a database system such relations, views, triggers, etc. as well as external components accessing a database system, such as queries extracted from procedures, object modules and their significant properties (e.g., attributes, conditions). Apart from the simple task of capturing the semantics of a database system, the graph model allows us to predict the impact of a change over the system and the application of graph-theoretic metrics. We furthermore study techniques and algorithms for handling changes occurring in the database schema, in such way that the human interaction is minimized. Thus, we provide a mechanism for enriching the evolution graph with evolution semantics such as evolution events and policies regulating its behaviour in the presence of hypothetical changes occurring in the database schema. Rules that dictate the proper actions, when additions, deletions or modifications are performed to relations, attributes and conditions (all treated as first-class citizens of the model) are provided. Specifically, assuming that a graph construct is annotated with a policy for a particular event (e.g., a relation node is tuned to deny deletions of its attributes), the proposed framework (a) performs the identification of the affected part of the graph and, (b) if the policy is appropriate, proposes the readjustment of the graph to fit to the new semantics imposed by the change. Additionally, we complement the proposed framework with a set of SQL extensions that allows the definition of evolution metadata with a feasible and efficient way. To this end, we employ graph theoretic and information theoretic properties of the evolution graph and establish a suitable set of measurements for evaluating the design quality of a database centric environment with respect to its ability to sustain evolution operations. All of the above concepts are implemented in a powerful and user friendly tool, called HECATAEUS, which is used for the application of the framework on real world evolution scenarios.