Report ID
1999-21
Report Authors
I. Stanoi, D. Agrawal, and A. El Abbadi
Report Date
Abstract
Data warehouses are designed mostly as centralized systems, and the majority ofupdating maintenance algorithms are tailored for this specific model.Maintenance methods have been proposed either under the assumption of a singleview data warehouse, a multi-view centralized model, or a multi-viewdistributed system with strict synchronization restrictions. We argue thatextending this model to a multi-view distributed one, is a practicalgeneralization of the data warehouse system, but also the base of a growingnumber of applications based on the idea of cooperative views. In thiscontext, existing maintenance protocols either solve the multi-viewsynchronization and consistency problem by grouping out of order updates inbatches, or by creating multiple versions of the data. In this paper wedevelop a general framework for modeling the maintenance of multi-views in adistributed, decentralized data warehouse, together with an efficientincremental algorithm. To our knowledge, there is no other proposal for amethod that incorporates individually and asynchronously updates to differentviews that are related through derivation dependencies.
Document
1999-21.ps322.46 KB