Emerging Database course Distributed Database Distributed Database When
Emerging Database course: Distributed Database
Distributed Database When the computer network technologies matured in the late 1980 s, distributed database concepts prevailed. Research focused on placing the existing central ized databases into networks so that these databases could utilize advantages provided by the networks, such as local autonomy, improved performance, improved reliability and availability, economics, expandability, and sharability [31]. Through the networks the processing logic, functions, data, and controls could be distributed to local processing nodes. By interconnecting the networks, the local processing elements could achieve an assigned task cooperatively. The design of a distributed database system can start from either a top down or a bottom-up approach. The top-down approach fragments the global schema into local schema, while the bottom-up approach emphasizes the inte gration of local conceptual schema.
The latter approach creates a multi database architecture that compromises the existing heterogeneous database architec tures. In this approach, the different local conceptual schema are transformed by a translator modular for schema integration. Since the local schema are dif ferent, the data of a local processing node are unable to be backed-up from one processing node to another in the multi database architecture. Fortunately, most approaches focus on the top-down approach while using the concepts of a relational database for the design phase. This approach has been successfully commercialized. The system designer must carefully select the process used to fragment a source table into small tables and replicate them in local processing nodes. The process used should be carefully combined with relational normalization. There are two kinds of fragmentation schema: vertical and horizontal fragmentation. The vertical fragmentation process cuts the source table into several fragments (tables) with different attributes. For joining purposes, the key attributes should be included in all fragments. The horizontal fragmentation slices the tuples of a source table into smaller tables with the same schema. The tuples in different fragments should not be overlapped.
Bachmann proposed the following set of criteria for classifying distributed databases even though the degree of distribution may vary greatly: 1. Degree of coupling of the processing nodes, e. g. , strong or weak. Interconnection structure of the nodes, e. g. , point-to-point. Independence of data components. Synchronization between data components. For increasing availability and reliability, some distributed databases maintain multiple copies in different sites. However, updating the data is a difficult job. For example, when a processor wants to update the data in one site, it needs to update all copies to maintain data consistency. The intercommunication among sites for granting the data locks and the permissions for data access increases the network loading. Related issues are the handling of updates when one site crashes at the moment of updating and finding the most current copy when recovering. Therefore, timestamping, locking, updating, recovering, and dead lock, weave a complex web for a distributed database. Also, the reliability of networks creates an obstacle in using a distributed database. Many questions, such as how many copies should be kept and where they should be kept, affect cost and performance. Too many replicates may decrease the system perfor mance since the network communication among nodes for maintaining the data consistency will increase.
REFERENCES • Timon C. Du. , Emerging Database System Architectures • Bochmann, G. Concepts for Distributed Systems Design • Capron, H. L. Computers: Tools for an Information Age
- Slides: 5