| 摘要 |
A protocol is defined to provide a method of controlling, storing, distributing, and recovering replicas among a group of databases sharing common data within a processing network. This method does not broadcast any pending update information to its "peer" databases and each "peer" database does not store or use any information about any other transaction that may be occurring on other databases. Tables within a commonly accessible and logically separate control database (or multiple databases) hold both replica control information and any replica for each table that will be propagated. In the preferred embodiment, replication daemons (or agents) handle the propagation and the application of the replica to all other "peer" databases. Update receipt information may be added to those "peer" databases that have been designated for recovery processing. When necessary, the update receipt information is used to restore the control information in the control database.
|
