Conflict Tolerant Queries in AURORA

Conflict tolerant queries are a new way of dealing with instance level conflicts in data integrated from multiple sources. In contrast to the traditional approach of resolving such conflicts during schema integration using aggregation functions, we establish a query model and processing techniques to tolerate these conflicts at query time to a degree specified by the users. Resolutions are only performed to produce conflict-free results. Currently we support 3 levels of conflict tolerance: HighConfidence, RandomEvidence, and PossibleAtAll and allow user-defined functions to be used for conflict resolution. The approach reduces the overhead of conflict detection and resolution and lends itself to new query optimization techniques. Fundamentally, our approach allows users to handle conflict at a coarse granularity to achieve better query performance when conflict resolution requirements are relaxed and when data contain occasional conflicts.