Reliable distributed database management

A reliable distributed database must manage dispersed and replicated data, making them available to users in spite of hardware failures. In this paper we study the algorithms and techniques that can achieve this reliability. Three scenarios are considered for reliable (i.e., nonpartitionable) networks. In the first, data are distributed but not replicated; in the second, data are replicated; and in the third, both data and processing are fully replicated in an n-modular redundancy strategy. A brief survey of methods for handling network partitions is included.