Quality-aware data management

Traditional data management systems operate over precise data and produce exact answers to posed queries. When data is imprecise, or when query processing is expensive, it may be costly or impossible to give perfect answers to applications. Quality-Aware Data Management attempts to deal with this problem by emphasizing (i) quality-performance tradeoffs that can be exploited to meet applications' quality requirements or minimize the use of resources, and (ii) quality guarantees about imprecise data and the approximate results produced by queries, allowing users to properly interpret them. This dissertation examines quality-aware data management techniques for approximate aggregate query processing over large datasets, for data capture and archival of highly dynamic data sources, for approximate selection queries over imprecise objects that can be probed to reveal their precise versions, and for multi-version selection queries.