Research And Implementation Of Synchronization Middleware For Heterogeneous Databases

With the wide use of databases and rapid development of network and distributed technology, to guarantee the consistency of distributed data becomes an urgent need for database development. But the characteristic of application and the employ circumstance make the choice diversity of database products. So it is necessary to design an infrastructural and generic platform to synchronize data in heterogeneous database systems.As commercial database synchronization products with unshared code use different standards, they have less flexibility and can not interconnect, intercommunicate or inter-operate. The solutions in academic literatures still fall short of efficiency, availability and flexibility. To solve these problems we design and implement an application independent middleware system for heterogeneous database synchronization. The main contributions of this thesis are summarized as follows.First, to improve the existing methods of data change capture which have little availability and high system resource occupancy, we propose log-table data change capture, a method of data change capture based on the trigger and the customized table. This method can be applied on most databases with less system resources occupancy meanwhile keeping the autonomy of client systems and the availability of data.Second, to achieve synchronization, we design a timestamp based optimistic synchronization strategy which is based on the log-table data change capture. This strategy only replicates the updated data and takes the advantage of both the state transfer replication and the operation transfer replication. Combined with concurrency control, it can stably and reliably synchronize data.Third, aiming at the heterogeneity of database platform, data type and data structure, we propose a method to transform heterogeneous data types into a uniform data type, and use XML based data format enlightened by the relational data model. We adopt the SyncML Protocol which is a widely used open sync protocol to transfer data, which enables our sync system to interconnect, intercommunicate and inter-operate with other SyncML Protocol based systems, and guarantees the openness and extensibility of the sync system.Fourth, to verify the solutions proposed in this paper, an available and configurable synchronization middleware system for heterogeneous databases is implemented. We test its performance of synchronization, analyze the results and seek the factors which affect the performance of our sync middleware system. Experimental results validate that this system can stably and reliably synchronize data in heterogeneous database systems meanwhile keeping the autonomy of client systems and the availability of data.