Don't Mind Your Vocabulary: Data Sharing Across Heterogeneous Peers

The strong dynamics of peer-to-peer networks, coupled with the diversity of peer vocabularies, makes query processing in peer database systems a very challenging task. In this paper, we propose a framework for translating expressive relational queries across heterogeneous peer databases. Our framework avoids an integrated global schema or centralized structures common to the involved peers. The cornerstone of our approach is the use of both syntax and instance level schema mappings that each peer constructs and shares with other peers. Based on this user provided mapping information, our algorithm applies generic translation rules to translate SQL queries. Our approach supports both query translation and propagation among the peers preserving the autonomy of individual peers. The proposal combines both syntax and instance level mappings into a more general framework for query translation across heterogeneous boundaries. We have developed a prototype as a query service layer wrapped around a basic service providing heterogeneity management. The prototype has been evaluated on a small peer-to-peer network to demonstrate the viability of the approach.