This discussion provides an overview of additional distributed relational database concepts.
On IBM® systems, some distributed relational database support is provided by the DB2 Universal Database™ for Linux®, UNIX® and Windows® and DataPropagator™ for iSeries™ products. In addition, you can use some of these concepts when writing iSeries application programs.
DB2 Universal Database™ for iSeries supports both the remote unit of work and distributed unit of work with Advanced Program-to-Program Communications (APPC) and TCP/IP communications, starting in OS/400® V5R1. A degree of processing sophistication beyond the distributed unit of work is a distributed request. This type of distributed relational database access enables a user or application program to issue a single SQL statement that can read or update data at multiple locations.
Tables in a distributed relational database do not have to differ from one another. Some tables can be exact or partial copies of one another. Extracts, snapshots, and replication are terms that describe types of copies using distributed processing.
Extracts are user-requested copies of tables. The copies are extracted from one database and loaded into another specified by the user. The unloading and loading process might be repeated periodically to obtain updated data. Extracts are most useful for one-time or infrequent occurrences, such as read-only copies of data that rarely changes.
Snapshots are read-only copies of tables that are automatically made by a server. The server refreshes these copies from the source table on a periodic basis specified by the user—perhaps daily, weekly, or monthly. Snapshots are most useful for locations that seek an automatic process for receiving updated information on a periodic basis.
Data replication means the server automatically updates copies of a table. It is similar to snapshots because copies of a table are stored at multiple locations. Data replication is most effective for situations that require high reliability and quick data retrieval with few updates.
Tables can also be split across computer servers in the network. Such a table is called a distributed table. Distributed tables are split either horizontally by rows or vertically by columns to provide easier local reference and storage. The columns of a vertically distributed table reside at various locations, as do the rows of a horizontally distributed table. At any location, the user still sees the table as if it were kept in a single location. Distributing tables is most effective when the request to access and update certain portions of the table come from the same location as those portions of the table.