1. A distributed database consists of two or more SERVERDBs which have a common catalog and common user data.
2. There is one system database administrator (SYSDBA) on each SERVERDB. The SYSDBA may drop all users of this SERVERDB, even those not created by him.
3. Each user is assigned one of these SERVERDBs as a HOME SERVERDB. The user data in the base tables that are owned by the user, as well as the partial catalog of the user, are always stored on the HOME SERVERDB of this user.
4. The catalog consists of one part that is copied to all SERVERDBs and of another part that is only stored on one SERVERDB.
5. The partial catalog stored on all SERVERDBs contains the definitions of SERVERDBs, users, and usergroups.
6. The partial catalog that is only stored on one SERVERDB comprises the partial catalogs of all users and usergroups for which this SERVERDB is the HOME SERVERDB. These partial catalogs describe all database objects defined by these users and usergroups, except for the set specified in item 5.
7. A table name specification does not contain any specification of the SERVERDB to which the table is assigned. SQL statements are independent of the SERVERDB to which a user or table is assigned. Each SQL statement can be executed from any SERVERDB as long as all SERVERDBs are in WARM mode and network communication between the SERVERDBs is possible. If one of these requirements is not met, the following conditions apply.
8. (Metadata) Data stored on one SERVERDB can only be modified when this SERVERDB is in WARM mode. If the Session of the user who wants to make these modifications was not started on the SERVERDB where the data to be modified is stored, the two SERVERDBs must be connected to each other within the network.
9. (Metadata) Data stored on all SERVERDBs can be modified even if not all SERVERDBs are in WARM mode or if network communication to some SERVERDBs is interrupted. SERVERDBs that are shut down or not accessible within the network are informed about modifications to the database as soon as they are put into WARM mode by using the Operating / Restart / Warm menu function of the ADABAS component CONTROL or when network communication has been reestablished.
10. Special processing is done if the network of SERVERDBs has split into two subnetworks which can no longer communicate with each other within the network. (Metadata) Data stored on a SERVERDB contained in one of the subnetworks can be modified from any SERVERDB belonging to that subnetwork. (Metadata) Data stored within the other subnetwork cannot be modified.
11. To prevent the two subnetworks from contradictory modifications to the replicated (metadata) data, ADABAS determines the subnetwork with the greater number (the so-called majority) of SERVERDBs within the whole network. The subnetwork containing the majority is then allowed to modify the (metadata) data. This procedure is called the majority concept. For two subnetworks of equal size, ADABAS decides the one that is to represent the majority. The minority subnetwork is not allowed to modify replicated data. In the case of read-accesses, it may happen that the minority subnetwork does not receive the latest state of (metadata) data updated by the majority. ADABAS displays warnings to inform the user about such a state.
12. Information about which SERVERDBs belong to the majority is contained in the corresponding system tables.