AWS DMS CDC polling frequency

From AWS Support engineer, there is no mecanism to reduce AWS DMS polled period.

When using AWS Database Migration Service (DMS) for Change Data Capture (CDC) with an Oracle 19c source instance, you can indeed adjust the polling frequency. However, the link you've mentioned doesn't directly relate to polling frequency settings.

For Oracle sources, AWS DMS offers two methods for reading redo logs during CDC: Oracle LogMiner and AWS DMS Binary Reader. By default, AWS DMS uses LogMiner, but for better performance, especially with Oracle RAC or when dealing with high volumes of changes, it's recommended to use Binary Reader.

To use Binary Reader, you need to set the following extra connection attributes:

useLogMinerReader=N;useBfile=Y

While these attributes don't directly control polling frequency, using Binary Reader can improve overall performance and potentially reduce the effective polling interval.

If you're experiencing a polling period of about 5 seconds and wish to lower it, you might want to consider adjusting the following settings:

1. CDC_LATENCY: This parameter controls how often DMS checks for new changes in the source database. The default is 30 seconds, but you can lower this value.

2. BATCH_APPLY: This setting determines how many events are grouped before being applied to the target. Lowering this value might increase the frequency of updates.

3. COMMIT_RATE: This controls how often transactions are committed on the target. Adjusting this might affect the apparent polling frequency.

These settings can be modified in the task settings of your DMS task. Keep in mind that very frequent polling can increase the load on your source database, so it's important to balance performance needs with system resources.

Additionally, ensure your source database is properly configured for CDC, including appropriate supplemental logging and retention of redo logs. If you're using Binary Reader with an Oracle RAC setup, you might also want to consider adjusting the parallelASMReadThreads and readAheadBlocks attributes for optimal performance.

Remember to thoroughly test any changes to these settings in a non-production environment before applying them to your production workload.

Sources
Use Oracle Real Application Clusters as a source for AWS DMS | AWS Database Blog
OracleSettings - AWS Database Migration Service
Filter transactions from users with AWS DMS from an Oracle database | AWS Database Blog