A user-centric view of data warehouse maintenance issues

Data in warehouses need to be updated in a timely manner from underlying operational data sources. This is referred to as warehouse maintenance. Not all of the data in the warehouse has the same requirement in terms of staleness (how old can it be with respect to the actual data), or its inverse freshness, and consistency (combining data from autonomous sources may give rise to some inconsistency). Given the requirements and schema information of a data warehouse, identifying policies for change detection and warehouse maintenance is a complex task. In this paper we identify a problem with current specification of user requirements, and suggest a specification scheme that is more general and user-oriented than extant suggestions. We also survey various policies that have been proposed for data propagation and analyse how change detection capabilities of sources influence user, as well as system requirements.