Constructing a temporal database management system

Abstract

Temporal database management systems provide integrated support for the storage and retrieval of time-varying information. Despite the extensive research which has been performed in this area over the last fifteen years, no commercial products exist and few viable prototypes have been constructed. It is our thesis that through the use of the proper abstractions, it is possible to construct a temporal database management system with robust semantics, without sacrificing performance, and with minimal implementation cost. Our approach parallels the development of relational database management systems, beginning with a theoretically sound abstract model, and then developing the underlying techniques to efficiently implement it. The major theme underlying this research is practicality, in terms of both semantics and implementation. We will show that expressive temporal semantics can be supported while still maintaining reasonable performance, and with relatively small implementation effort. This is made possible, in part, by minimally extending the relational model to support time, thereby allowing the reuse or easy adaptation of well-established relational technology. In particular, we investigate how relational database design, algebras, architectures, and query evaluation can be adapted or extended to the temporal context. Our aim is that software vendors could incorporate these results into existing non-temporal, commercial products with relatively small effort.