Efficient query reformulation in peer data management systems

  • Authors:

  • Igor Tatarinov;Alon Halevy

  • Affiliations:

  • University of Washington, Seattle, WA;University of Washington, Seattle, WA

  • Venue:
  • SIGMOD '04 Proceedings of the 2004 ACM SIGMOD international conference on Management of data
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Peer data management systems (PDMS) offer a flexible architecture for decentralized data sharing. In a PDMS, every peer is associated with a schema that represents the peer's domain of interest, and semantic relationships between peers are provided locally between pairs (or small sets) of peers. By traversing semantic paths of mappings, a query over one peer can obtain relevant data from any reachable peer in the network. Semantic paths are traversed by reformulating queries at a peer into queries on its neighbors.Naively following semantic paths is highly inefficient in practice. We describe several techniques for optimizing the reformulation process in a PDMS and validate their effectiveness using real-life data sets. In particular, we develop techniques for pruning paths in the reformulation process and for minimizing the reformulated queries as they are created. In addition, we consider the effect of the strategy we use to search through the space of reformulations. Finally, we show that pre-computing semantic paths in a PDMS can greatly improve the efficiency of the reformulation process. Together, all of these techniques form a basis for scalable query reformulation in PDMS.To enable our optimizations, we developed practical algorithms, of independent interest, for checking containment and minimization of XML queries, and for composing XML mappings.