Propagating Integrity Information among Interrelated Databases

Data integrity policies often require that quality and integrity metadata be generated and communicated to potential users. However, in data warehouses, federations, and other multi-tier databases, administrators at different tiers use different schemas. Data at the upper tier is derived from the lower, so we consider the upper tier’s tables to be views, derived by SQL-like expressions. Unfortunately, an assertion about some granule in the sources (a table, column, or cell) is often meaningless to view users, and vice versa. An understanding of the SQL view gives intuitive guidance for propagating such metadata, but not explicit semantics. It appears feasible to create a system that drastically reduces the skill and labor required for propagating metadata and events between the tiers. We show many examples where, based on the view query and the metadata on the relevant sources, one can automatically generate useful propagation rules. Propagation downward from views to sources is also handled. Our approach is to automate the easy cases (which we expect to be quite common), and to assist on harder cases. Knowledge of specific metadata types or query operators can be supplied incrementally. If the view’s query expression is difficult, one may compose the propagation rules from its constituent operators.