Transaction Processing in Peer Database

Lianxiang Qiu

Abstract

In a peer-to-peer database network, each peer is equipped with a relational database. There is no global schema and data vocabulary in the system. Each peer manages its own data and only shares relevant data with acquainted peers. In the Hyperion project, mapping tables are used to bridge the differences. To support data sharing, the use of mapping tables has been proposed. These mapping tables use pairs of corresponding data values that reside in different peers. A framework has been proposed to answer user’s query, that is, the local query is answered locally first, then the query is translated, and propagated to other acquainted peers according to the mapping tables, so the user can get the answers from the whole network instead of just from his/her own peer. Another important issue for P2P databases is the transaction management. Update to the database is initiated by a local peer without any execution control of other peer databases. The update is executed locally, then propagated to other acquainted peers and execute remotely. We introduce a mechanism to ensure the correct execution of concurrent transactions over the P2P network.