Research On And Implementation Of An Open Framework For Heterogeneous Database Replication

Database replication is one of the key techniques in distributed environments. Because the database heterogeneities become more and more popular, the researching on heterogeneous database replication is very important both theoretically and practically.As the beginning of whole work, the thesis analyzes and summarizes the key methods and techniques of replication implementation in the aspects of architecture, capturing, distribution, synchronization, and conflicts detection & resolution.According to their basic principles and methodologies, the thesis classifies the current heterogeneous database replication solutions as three modes:Database Vendor Mode, Replication Middleware Mode and Replication Engine Mode, and summaries both the advantages and disadvantages.Beginning with categorizing the heterogeneity forms, the thesis proposes an open framework 234-5 in a creative way. The framework firstly abstracts two types of replication objects: Publication and Subscription; then constructs a three-layer open architecture: Connection Layer, Implementation Layer and Application Layer; Four components (Capturing Manager, Distributor, Synchronizer and Conflicts Manger) are defined in the Implementation Layer; Thus it can entirely solve five problems: Fully Heterogeneous, All-Round functions, Transplantability, Expandability and Global Performance.Aiming at the Multi Replication Definition, the thesis presents three optimized propagation algorithms: D-M, ILS and LIS. D-M Algorithm gets the optimized individual cost per sites pair, whereas ILS and LIS Algorithm reduce the total costs to minimum respectively in chain topology and other conditions. They are proved to be correct and effective both theoretically and experimentally.The thesis proposes an MLS-based completive replication authorization management (CRAM) mechanism. CRAM appropriately specifies the objects of authorization management, and greatly expands the operation privileges and their status, the influences to performance are tested at last.According to Framework 234-5 and other theories, a prototype named Cabinet Replicator is developed. It consists of PCL Subsystem, DB2-I Subsystem, VFP-I Subsystem, an administration GUI tool, and several example applications. The testing results show that Cabinet Replicator actualizes the characters of above theories, and achieves our expecting.