# Change Data Capture

With Change Data Capture (CDC), you can track changes in table data. YDB provides access to changefeeds so that data consumers can monitor changes in near real time.

## Enabling and disabling CDC

CDC is represented as a data schema object: a changefeed that can be added to a table or deleted from them using the ADD CHANGEFEED and DROP CHANGEFEED directives of the YQL ALTER TABLE statement.

## Reading data from a topic

You can read data using an SDK or the YDB CLI. As with any other data schema object, you can access a changefeed using its path that has the following format:

path/to/table/changefeed_name


For example, if a table named table contains a changefeed named updates_feed in the my directory, its path looks as follows:

my/table/updates_feed


Before reading data, add a consumer. Below is a sample command that adds a consumer named my_consumer to the updates_feed changefeed of the table table in the my directory:

ydb topic consumer add \
--consumer-name=my_consumer


Next, you can use the created consumer to start tracking changes. Below is a sample command for tracking data changes in the CLI:

ydb topic read \
--consumer=my_consumer \
--format=newline-delimited \
--wait


## Impact on table write performance

When writing data to a table with CDC enabled, there are additional overheads for the following operations:

• Making records and saving them to a changefeed.
• Storing records in a changefeed.
• In some modes (such as OLD_IMAGE and NEW_AND_OLD_IMAGES), data needs to be pre-fetched even if a user query doesn't require this.

As a result, queries may take longer to execute and size limits for stored data may be exceeded.

In real-world use cases, enabling CDC has virtually no impact on the query execution time (whatever the mode), since almost all data required for making records is stored in the cache , while the records themselves are sent to a topic asynchronously. However, record delivery background activity slightly (by 1% to 10%) increases CPU utilization.

When creating a changefeed for a table, the number of partitions of its storage (topic) is determined based on the current number of table partitions. If the number of source table partitions changes significantly (for example, after uploading a large amount of data or as a result of intensive accesses), an imbalance occurs between the table partitions and the topic partitions. This imbalance can also result in longer execution time for queries to modify data in the table or in unnecessary storage overheads for the changefeed. You can recreate the changefeed to correct the imbalance.

As a load generator, you can use the feature of emulating an online store built into the YDB CLI:

1. Initialize a test.

ALTER TABLE orders ADD CHANGEFEED updates WITH (
FORMAT = 'JSON',
);

3. Create a consumer:

ydb topic consumer add \
--consumer-name=my_consumer

4. Start tracking changes:

ydb topic read \
--consumer=my_consumer \
--format=newline-delimited \
--wait

...