Composing dimension and fact mappings in peer data warehouses

Abstract

Semantic mappings are correspondences that are established between instance or schema level vocabularies of autonomous and heterogeneous data sources. The importance of semantic mappings is steadily increasing in recent data sharing architectures as these mappings enable query answering across heterogeneous boundaries. Peer data management systems (PDBMSs) are one such architecture that has semantic mappings at its core. These systems are made up of fully autonomous network nodes, called peers, which contain data sources to be shared with other peers, called acquaintances. A peer data warehouse (PDW) is a traditional data warehouse (with appropriate dimensions and facts) which is associated with a peer and managed by a PDBMS. We investigate the problem of composing semantic mappings between dimensions and facts of acquainted PDWs. Moreover, we give a distributed algorithm for composing these mappings expressed in the mapping language LAV (Local-As-View). Furthermore, we show that the complexity of the algorithm is quadratic and prove the correctness of the latter. Finally, we conduct several experiments that illustrate the robustness of the algorithm.