| With the impact of global wave of information and the constant expansion of computer applications, all walks of life deal with greater amount of information, particularly in the aerospace, communications, marine development, weather forecasting, military and other fields.T-he scale of data has reached TB class or even PB class .The traditional serial database system encountering such large-scale data has become unable to deal with. Therefore, it has become the consensus of the database fields to design the parallel database system used to store and manage large-scale data and to provide better performance. Parallel database system aims at high performance, high availability and high scalability to fully utilize multiprocessor platforms, working with a number of synergistic processors in order to achieve a faster response and analysis capabilities. At present, there have been some relatively sophisticated parallel database technology, but also there are some commercial parallel database products.With the previous research results, based on Cluster system of Shared-Nothing (SN) structure, the author proposed to develop the middleware software on the basis of MPI and RDBMS. The role of middleware is to shield parallel programming and database operations to the upper parallel database operations. Under the guidance of this thought, a parallel database prototype system PDBS-0 is formed. The main research results are as follows:1,A set of software and hardware research platform has been set up on the basis of theclusters.2,The theoretical framework of the parallel database system middleware software PDBS-0has been Put forward on the basis of MPI and Access.3,A lot of parallel database query operation modules have been achieved. Such as parallelsort, parallel selection and parallel join.4,A comprehensive test has been conducted and a detailed analysis of test results havebeen carried out; the key points of improving the performance of the whole system havebeen summerized. |
PDF Full Text Request