Massive parallelism for query answering in weakly integrated P2P systems

A new set of requirements for query processing has emerged, as Internet and Web-based query systems have become more prevalent. In this emerging data management domain, queries are posed over multiple semistructured information sources distributed across a wide-area network. Each source may be autonomous and may potentially have data of a different format and new sources are frequently added. We present the basic elements of a new weakly coupled semantics that correctly captures the modular structure of a data-intensive P2P system, and opens up the possibility of effective query answering techniques. One of the basic characteristics of peers in our approach is that they are based on abstract object types, which encapsulate a data integration system in order to deal with incomplete Web-based information, and to offer a rich ontology interface to their clients. We describe an incremental algorithm for answering conjunctive queries posed to one peer. The contributions to the query answer of any peer are given by the certain answers that such peer provides. We show that the algorithm supports the high parallelism in the process of query answering from different peers. Such parallelism can be modulated from the massive parallelism to the minor forms of parallelism, which depends from the user incremental-answering strategy.