Database design and translation for multiple data models

Database experts have debated over which model of data organization is best under various criteria. We believe that no model is clearly best and that future systems must support more than one data model. This dissertation explores methods and techniques for constructing a heterogeneous database management system from existing database systems. Logical and physical database design is studied in an environment which supports multiple data models and systems. Techniques are formulated to translate a database to an equivalent organization under a different data model and to convert the query portions of programs. A data model is developed to capture the semantic interrelationships supported by a database. The purpose of a logical design is to derive a schema which is well-behaved under the update operations of a particular model. A database specification in the semantic data model is mapped into the constructs of a target model, while preserving design goals related to the desirable behavior of a schema under update. The semantic data model is augmented with logical access paths to represent how semantic objects are interconnected. Physical storage structures are characterized by a small number of basic properties. The physical design method proceeds in two phases. Properties are first assigned to the logical access paths. Then, the logical access paths are implemented by choosing storage structures which support the assigned properties. Data translation is accomplished by recognizing constructs within the source database that correspond to a semantic object, and then mapping these into an equivalent realization in the target model. The logical access paths supported in the physical database direct the program translation. A sequence of low level operations are identified as a semantic access. A composed sequence of these accesses is mapped into a single high level query specification. The inverse translation is accomplished by mapping a high level specification into a sequence of low level operations which take advantage of the access paths efficiently supported by the database. A semantic data model was used to integrate the processes of database design and translation in an environment which supports multiple data models. The methodology for logical and physical database design presented in this dissertation can be used as the basis for an automated database design tool. The techniques for data and program translation make it possible to migrate data between different data models and systems without the need to reprogram user applications. These methodologies and techniques are needed to implement a heterogeneous system on top of existing systems.