Models of a very large distributed database

摘要

The problems inherent in managing a database distributed over a very large number of sites are considered. The applications of such databases to the provision of telecommunications and other public services are discussed. It is shown that the distribution and maintenance of the directory information describing object locations poses some fundamental problems. A new partially informed class of distributed databases is described which distributes the directory information on a "needs-to-know" basis. The class is described by models of the network topology, and by the knowledge available to each site. These proposals are sufficiently general to support the partitioning of data relations into distribution fragments, and for those fragments to be replicated at multiple sites.