The Extract process is configured to run against the source database, capturing data generated in the true source database located somewhere else. This process is the extraction or the data capture mechanism of Oracle GoldenGate.

You can configure an Extract for the following use cases:

• Initial Loads: When you set up Oracle GoldenGate for initial loads, the Extract process captures the current, static set of data directly from the source objects. This configuration of Extract process uses source tables as the source to capture data.

• Online Extract (Change Synchronization): When you set up Oracle GoldenGate to keep the source data synchronized with another set of data, the Extract process captures the DML and DDL operations performed on the configured objects after the initial synchronization has taken place. Extracts can run locally (upstream) on the same server as the database or on another server using the downstream integrated Extract (in case of Oracle database) for reduced overhead. It stores these operations until it receives commit records or rollbacks for the transactions that contain them. If it receives a rollback, it discards the operations for that transaction. If it receives a commit, it persists the transaction to disk in a series of files called a trail, where it is queued for propagation to the target system. All the operations in each transaction are written to the trail as a sequentially organized transaction unit and are in the order in which they were committed to the database (commit sequence order). This design ensures both speed and data integrity. This configuration of the Extract process uses database recovery logs or transaction logs as the data source. While capturing from the logs, the actual method varies depending on the database type. An example of this source type is the Oracle database redo logs, which are used for supplemental logging.

For a remote deployment, the source database and Oracle GoldenGate are installed on separate servers. Remote deployments are the only option available for supporting cloud databases, such as Azure for PostgreSQL or Amazon Aurora PostgreSQL.

For remote deployments, operating system endianness between the database server and Oracle GoldenGate server need to be the same.

Server time and time zones of the Oracle GoldenGate server should be synchronized with that of the database server. If this is not possible, then positioning of an Extract when creating or altering one will need to be done by LSN.

In remote capture use cases, using SQLEXEC may introduce additional latency, as the SQLEXEC operation must be done serially for each record that the Extract processes. If special filtering that would require a SQLEXEC is done by a remote hub Extract and the performance impact is too severe, it may become necessary to move the Extract process closer to the source database.

With remote deployments, low network latency is important, and it is recommended that the network latency between the Oracle GoldenGate server and the source database server be less than 1 millisecond.