Evaluating maintenance policies for externally materialised multi-source views

  • Authors:
  • Henrik Engström;Brian Lings

  • Affiliations:
  • Department of Computer Science, University of Skövde, Sweden;Department of Computer Science, University of Exeter, UK

  • Venue:
  • BNCOD'03 Proceedings of the 20th British national conference on Databases
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

In many applications data from distributed, autonomous, and heterogeneous sources need to be imported and materialised in a (client) system external to those sources. As changes are committed in the sources, the externally materialised view must be updated to reflect those changes. A maintenance policy determines when and how to conduct updates. As sources may not be cooperating maintenance of externally materialised views is different from traditional view maintenance. Previous studies on maintenance of externally materialised views have been heavily focused on algorithms that ensure view consistency. There are, however, other aspects of maintenance that, when considered, can affect choice of consistency algorithm. If, for example, auxiliary views are maintained in the view client it is possible to ensure strong consistency without complex algorithms. In our previous work we have studied how to select a maintenance policy for a single source view. In this paper we extend the work to evaluating maintenance policies for externally materialised views based on several sources. We explore views that are defined as the join of two independent sources, identifying the solution space in terms of possible policies, their implications for consistency and their required source capabilities. We use a testbed system to evaluate policy performance. The work confirms that the earlier results on single source maintenance extend to the multisource situation. In addition we show that the consistency preserving algorithms suggested in the literature are not always required. Actually, in all situations explored it has proved more efficient to use auxiliary views than policies which requires consistency preserving algorithms.