Research On Technology Of Data Mapping Inference In Peer Data Management Systems

Along with the emergence of more and more P2P softwares, P2P draws more and more attention in recent years. A peer data management system (PDMS) is a kind of P2P model appropriate for structured data sharing in which data mapping tables are used extensively during query process. But maping tables are typically created by domain specialists. With the goal of improving the level of data sharing in PDMS, the automatic tool for creating mapping table is researched thoroughly. The current situation of the data mapping inference problem in PDMS is analyzed in this paper, and the problem is discussed from a completely new perspective.Firstly, through thorough research, we found that existing data mapping inference algorithm contains some redundancy operations, and has not taken full potential of distributed computing and autonomy of peers. To resolve theses problems, some improving measures are proposed.Secondly, a dynamic data mapping inference algorithm is proposed. Existing algorithm inputs a set of mapping tables. When one of the mapping tables updates, it must be executed again in whole path regardless of plenty of correct middle operations and results. The dynamic algorithm which can avoid these repeating computing is given and its correctness is proved in this paper.Thirdly, data maping inference system is designed. Existing mapping inference algorithm considers inference problem and consistent problem of mapping tables from mapping constraints. We address a practical data mapping inference problem from users'requirements in this paper. To resolve the problem, a semantic graph model is proposed, based on which a data mapping inference system is designed.Finally, on the foundation of existing success peer data managementsystems, we provide a new peer data management system model in which we can evaluate our research. It can satisfy our special requirements. We validate research results in it.