From an operational perspective, OLTP systems represent a company's electronic central nervous system, so the databases that support these systems must be highly available. Oracle has a number of features that contribute to high availability:
Standby database
Oracle provides database redundancy by maintaining a copy of the primary database on another machine, usually at another site. Redo logs from the primary server are shipped to the standby server and applied there to duplicate the production activity. Oracle8i introduced the automated shipping of redo logs to the standby site and the ability to open the standby database for read-only access for reporting.
Oracle9i Release 2 introduced the concept of logical standby. With a logical standby database the changes are propagated with SQL statements, rather than redo logs, which allows the logical standby database to be used for other database operations.
Transparent Application Failover (TAF)
TAF is a programming interface that enables you to automatically reconnect a user session to another Oracle instance should the primary instance fail.
Advanced Queuing (AQ)
AQ provides a method for asynchronous, or deferred, intersystem communication, allowing systems to operate more independently. Avoiding direct system dependencies can help to avoid "cascading" failures, allowing interconnected systems to continue to operate even if one system fails. AQ is described in more detail in the following section.
Replication
You can use Oracle's built-in replication functionality to provide data redundancy. Changes made by transactions are replicated synchronously or asynchronously to other databases. If the primary database fails, the data is available from the other databases. As of Oracle9i Release 2, log-based replication and Advanced Queueing were included as part of Streams.
Real Application Clusters
Real Application Clusters (Oracle Parallel Server prior to Oracle9i), increased the scalability of the Oracle database. However, by supporting multiple instances with full access to one database, RAC also provides the highest levels of availability for protection from the failure of a node in a clustered or MPP environment. If one node fails, the surviving nodes provide continued access to the database. With Oracle Database 10g, grid computing deployment further extends availability capabilities.
For a more detailed discussion of high availability, see Chapter 10.