Concurrency control and data replication strategies for large-scale and wide-distributed databases

The explosion of the Internet and the advances in high-speed wide-area networks (WANs) provide the opportunity to distribute large database systems over geographically distant sites to improve the performance and the robustness of such systems. The traditional concurrency control and commitment control protocols should be re-designed to fit this new scenario. Because of the diversity of applications, it is possible that the correctness of a transaction and database consistency can be relaxed based on the requirements of the application. This paper focuses on developing revised protocols to provide different levels of consistency and performance to meet the different needs of various applications in large and widely distributed database systems. In our approach, the entire system is organized in a tree framework. The concepts of insular consistency and regional consistency are introduced to relax strict consistency from the time and space dimensions. A concurrency control protocol, called the Hierarchical Token Forwarding Protocol (HTFP), and a commitment control protocol, called the Hierarchical Asynchronized Multi-level-consistency Protocol (HAMP), are proposed to achieve different kinds of consistency in a single distributed database system.