Abstract
Data Grids allow for seeing heterogeneous, distributed, and dynamic informational resources as if they were a uniform, stable, secure, and reliable database. According to this view, current proposals for data integration on Grids are based on the notion of global schema built over a collection of autonomous information sources. On the other hand, in dynamic and distributed environments, such a hierarchical and centralized architecture is not well suited for effective information integration. Peer-to-peer data integration aims at overcoming these drawbacks by modeling autonomous information systems as peers, and establishing mappings among peers without resorting to any hierarchical structure. In this paper, we present Hyper, a joint research initiative of Università di Roma " La Sapienza " and IBM Italia, which aims at developing principles and techniques for peer-to-peer data integration on a Grid infrastructure. The main contributions presented are a semantic characterization of P2P data integration, the deployment of our P2P framework on a Grid architecture , and the design of a query answering algorithm that is coherent both with the semantics and with the Grid infrastructure.