Atomicity Implementation in Multidatabases with High Performance and Availability

摘要

Data replication may improve both performance, response time and data availability of a distributed system. In this paper we will generalize the basic backup/replication methods in order to design databases with high performance and availability. Most of the basic backup/replication methods have only semantic ACID properties, i.e. from an application point of view, the system should function as if all the traditional ACID properties were implemented. Therefore, the replication methods must be integrated by the so-called extended transaction models, where it is the designers' responsibility to lay down the level of the ACID properties. In this paper we will also describe a new extended nested transaction model, and we will illustrate how the model may be used to increase both performance and availability of a multidatabase by using the generalized versions of the basic backup/replication methods. In this paper the main objective will be to describe in detail how the global semantic atomicity property may be implemented on top of client/server technology in order to preserve high performance and data availability. How to implement the other global semantic ACID properties is described in Frank and Zahle(1998).