Multi-database Data Integration Platform Design

With the development of network and database, some corporations reformed themselves using Information Technology. Every department continues to build their data center. They chose hardware and software products according to their own needs. Many application systems had been deployed on the data center to improve office efficiency. After that it becomes very difficult to share data between different departments because of the difference of Data Model and System Architecture. Database vendors had provided some tools to solve this problem. But we just can use these tools in special scenerio or in some applications built on given products.Some scholars have done lots of researches, and they had raised some frameworks to screen the difference of System Architecture and Data Model between LDBS (local database manage system). The best framework is MDBS (Multi- Database System). If system is built using this framework, users will be able to query the data through a unified interface which MDBS provides.This paper presents a MDBS program in a number of Relational Database Management Systems environment. The program uses Middleware Architecture (Wrapper/Mediator). We just focus on the sharing of data. So we will not think about anything about database transaction processing and data consistency. First of all, we have put forward a model of Model Integration and XML Description. The model is based on the structure of table. It is easy for developers. After that we analyzed every part of Query Processing, such as Query Pretreatment, Query Decomposition, Data Integration and Management, Query Optimization. Global query will be changed equivalently in the process of Query Pretreatment. Its purpose is to accelerate Query Decomposition. It does that by tidying global query, combining query and changing query limitation expression. Because local DBMS just can execute the query which be written using its query language, we can not pass the global query to it directly. Before passing query to it we must decompose the global query. If we want to fetch data related to this global query, we must generate the query that can execute in the local DBMS and fetch data we need. In order to do that we will map the global fields to local fields and transfer query limitation expression to local query limitation expression according to the information of Mode Integration. The global query results will be generated by assembling data fragments that fetched by the local query in the process of Data Integration. Some methods have been introduced to improve the efficiency of query executing. I have given optimized algorithm especially under the circumstance of data replication.Finally this paper shows some other designs in detail and explains every module's function. We have done some tests to test this system. It can screen the difference of System Architecture and Data Model between LDBS completely. Now we can fetch data from MDBS directly just like MDBS have saved all data.