南宫NG·28

1. When the replication rule is initiated, the incremental module begins capturing increments according to the rule.
2. The full-quantity module starts acquiring information about the objects involved in the rule as set.
3. Begin acquiring table data according to concurrency settings and multiple queues. The data acquisition process is as follows:
a. Obtain the current transaction ID, such as SCN, LSN.
b. Conduct a full table scan and directly deliver the scan results to the destination host.
4. The full-quantity loading module will load the data into the destination database according to the acquired concurrency.
5. After the data in the table is loaded, the software begins creating indexes, constraints, and other objects.
6. When all objects and data are loaded.

1. The reading (iareader) thread captures changes in the redo log and transmits these changes to the analysis (analyzer) thread.
2. The analysis thread generates TF files for DML and DDL changes to be replicated based on rule settings. TF (transaction file) records historical transactions.
3. The sender thread sends changes in the TF file to the disk of the target host over the network.
4. The queue allocation thread reads the TF file and automatically allocates different queues based on table dependencies.
5. The loading (apply) thread loads the operations into the target database.
The capture process collects changed data, and the data capture process brings almost no overhead to the production database. The capture process reads information from the database's pre-write log files, so the replication process does not cause significant performance issues for the production database instance. Only changed data is transmitted, not all information in the transaction log files, so the network load is relatively low. Although a small number of objects need to be installed in the source database to store some basic information about replication, the source database does not need to participate in the data capture and transmission process. The capture process can read both online log files and archived log files. This capability greatly enhances system robustness. For example, if the capture process is stopped for some reason, data synchronization will automatically continue from the last termination point when it restarts, and replication will not be affected.