Semantic data sharing with a peer data management system

Data sharing is a ubiquitous problem. Large enterprises, governments, research communities, and people with common interests are interested in making their data available as well as accessing others' data. The World Wide Web (WWW) and Peer-to-Peer (P2P) networks have been quite successful in providing an easy way for users to share their data. A serious limitation of these approaches is that they support keyword queries only. Data Integration and Data Warehousing are two common approaches that support database-style queries. These technologies are rather heavyweight, however. They impose a centralized mediated schema, which impedes schema evolution and complicates the data sharing process. Peer Data Management have recently been proposed as a more flexible approach to semantic data sharing. A Peer Data Management System (PDMS) uses a semantic network of mappings between peer schemas to reformulate user queries over every peer in the network. This dissertation proposes a number of techniques for scalable data sharing using a PDMS. The proposed techniques have been implemented in the Piazza PDMS and evaluated using two real-life data sets. The techniques are based on novel algorithms for reasoning about XML queries: XML query containment, minimization and rewriting, as well as XML mapping composition. These algorithms are essential in many applications outside the scope of PDMSs.