Building a system for replication means addressing a variety of design and configuration issues that are irrelevant in a standalone environment. As a general rule, it is not possible to “turn on” replication for an existing, standalone system, though some have tried. Items requiring attention include replication architecture, data consistency, data extraction, schema differences, primary key constraints, and confliction avoidance—to name a few. The sections that follow highlight these concerns.
There are two broad categories of replication architecture: log-based replication and transactional replication. Log-based replication works by examining the database’s transaction logs (redo logs in the case of Oracle) and forwards committed changes to other participating databases as needed. (Quest Software’s Shareplex product is an example.) Transactional replication, on the other hand, works by either querying the remote database (in the case of read-only snapshots) and adding triggers to replicated tables that effect the forwarding of changes to remote sites. Oracle’s advanced replication facilities are an example of this technology. One of your first choices is which of these two mechanisms to use.
Log-based replication has various advantages and disadvantages.
The primary benefits of log-based replication are speed and ease of configuration. Since this technology simply forwards changes from the database ...