An application of fragmentation transparency in a distributed database system: A case study

Distributed database theory for a single view of a distributed database consisting of homogeneous relational databases has been known for over a decade. The application of some of this theory in commercial products has not been established. Although some commercial distributed database products support naming and location transparency, these products do not provide full support for fragmentation transparency. Providing support for declaring and reasoning about fragments requires adding data definition capabilities and updating complex query optimizers. This article presents a syntax for the specification of fragmentation and describes a prototype for reasoning about the fragmentation criteria. This prototype examines the feasibility of not having to update the existing query optimizer by taking a distributed SQL query and fragmentation specification and generating a sequence of SQL queries expressed over the fragments, which is then input to the established query optimizer. This article provides a case study of this approach on a commercial distributed database product. A bench-mark for distributed databases for decision support, known as D³S, is introduced and used in the case study. The study shows that a two-step process of first optimizing with respect to the fragmentation criteria, followed by the existing distributed query optimization, results in performant query evaluation plans.