System P: Completeness-driven Query Answering in Peer Data Management Systems

Abstract: Peer data management systems (PDMS) are a highly dynamic, decentralized infrastructure for large-scale data integration. They consist of a dynamic set of autonomous peers inter-connected with a network of schema mappings. Queries submitted at a peer are answered with local data and by data that is reached along paths of mappings. Due to redundancies in the mapping network, query answering in PDMS can be very inefficient if the complete query result is to be computed. System P, a fully functional PDMS, compromises the completeness of the query result and reduces cost by pruning the query plan at mappings that are estimated to yield only few result tuples. The demo illustrates the following main components of System P: (1) adaptive estimation of result cardinalities of intermediate queries using histograms, (2) completeness-driven query planning under limited resources using specialized heuristics, and (3) the automatic generation of heterogeneous PDMS test instances, controll...

• Type: Conference
• Year: 2007
• Where: BTW
• Authors: Armin Roth, Felix Naumann
• Keywords: BTW 2007, Data Management Systems, Large-scale Data Integration, Peer Data Management