Proposition of the Temporal Variation Data model and Evaluation of an Implementation
Date of Award
Doctor of Philosophy (PhD)
Graduate School of Computer and Information Sciences
This dissertation identifies the need to develop practical implementations of temporal databases. Most databases today model the real world at one particular time. As new information is collected older information is deleted. This degrades the capabilities of decision support systems because only trends that were anticipated and included in the design of the database can be reported. The retention of older information requires significant increases in resources. This dissertation focuses on temporal databases which retain older information rather than delete it.
This dissertation reviews the relevance and significance of the use of temporal databases. It then presents a problem that has been identified through review of the literature on temporal databases. Using current models to design relations in a temporal database results in large databases that are costly to query. The Temporal Variation Data Model (TVDM), is proposed to address this problem. There are two basic approaches to implementing temporal databases. These are attribute and tuple versioning. Ahn (1986) provides calculations to compare the storage requirements between these alternatives. Shiftan (1986) assesses the temporal differentiation of attributes as an implementation strategy. His proposed implementation included both tuple and attribute versioned relations. Shiftan notes an additional step, separating event information, for use in organizing tuples. That additional step, taken concurrently with additional consideration of the variation of those attributes Shiftan included in tuple versioned relations, serves as the foundation for the TVDM.
Using one set of source data, attribute and tuple versioned databases are created, then the TVDM is used to create a third temporal database. This dissertation reports the results of a case study that includes a three-way comparison. It found increased utility for a temporal database as a result of using the new data model. Utility is operationally defined as a scaled comparison of central processing unit (CPU) time, input/output activity (VO), and required storage measurements. The CPU time and VO requirements are recorded for a series of test queries accessing each of the alternatives. The TVDM based database in the case study required the least storage space and had the lowest requirements for CPU time and VO.
Mark A. Brown. 1998. Proposition of the Temporal Variation Data model and Evaluation of an Implementation. Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, Graduate School of Computer and Information Sciences. (426)