On using historical update information for instance identification in federated databases

To support database interoperability in federated databases systems, it is critical to be able to identify (potentially) equivalent data instances from individual autonomous database components. Since the components in a federation are autonomous, their data may be updated asynchronously, viz., modifications to a real world entity may be captured in different databases at different times; the authors term this effect update heterogeneity. Existing approaches largely base data instance similarity identification only on current attribute/property values; in the face of update heterogeneity, this is inadequate. They present an approach to address the problem of update heterogeneity in the federated databases context. They employ a probabilistic model, which utilizes historical database update information to estimate the degree of similarity between candidate data instances from different database components. They employ transaction history (log) information to this end, which is typically already available in the component database systems. They have experimentally implemented and tested this approach within the context of a prototype experimental federated databases system, FeXpress.