DCU charges for each feature are calculated as follows:
- Data quality
Auto data quality scans:
- The DCU-hour consumption is proportional to the processing involved in profiling the data and computing the data quality metrics. This is billed per second, with a minimum of one minute.
- The charge depends on the number of rows, the number of columns, the amount of data that you've scanned, the data quality rule configuration, the partitioning and clustering settings on the table, and the frequency of the scan.
- For data quality anomaly detection scans, DCU charges do not apply. Instead, standard BigQuery pricing for compute, storage, and BQML model training, processing, and deploying apply.
- Using a custom execution identity changes how you are billed for the scan. When you specify a custom execution identity, the compute and storage costs associated with the scan are billed directly to your BigQuery project, bypassing the standard Dataplex Premium SKUs.
There are several options to reduce the cost of auto data quality scans:
To filter aggregate charges, use the following labels available in billing export in BigQuery:
- goog-dataplex-datascan-data-source-dataplex-entity
- goog-dataplex-datascan-data-source-dataplex-lake
- goog-dataplex-datascan-data-source-dataplex-zone
- goog-dataplex-datascan-data-source-project
- goog-dataplex-datascan-data-source-region
- goog-dataplex-datascan-id
- goog-dataplex-datascan-job-id
2. Data profiling
Data profiling scans:
- The DCU-hour consumption is proportional to the processing involved in profiling the data and computing the data quality metrics. This is billed per second, with a minimum of one minute.
- The charge depends on the number of rows, numbers of columns, the amount of data scanned, partitioning and clustering settings on the table, and the frequency of the scan.
- Using a custom execution identity changes how you are billed for the scan. When you specify a custom execution identity, the compute and storage costs associated with the scan are billed directly to your BigQuery project, bypassing the standard Dataplex Premium SKUs.
There are several options to reduce the cost of data profiling scans:
- Sampling
- Incremental scans
- Column filtering
- Row filtering
- To separate data profiling charges from other charges in the Dataplex premium processing SKU, on the Cloud Billing report, use the label goog-dataplex-workload-type with value DATA_PROFILE
To filter aggregate charges, use the following labels available in billing export in BigQuery:
- goog-dataplex-datascan-data-source-dataplex-entity
- goog-dataplex-datascan-data-source-dataplex-lake
- goog-dataplex-datascan-data-source-dataplex-zone
- goog-dataplex-datascan-data-source-project
- goog-dataplex-datascan-data-source-region
- goog-dataplex-datascan-id
- goog-dataplex-datascan-job-id
3. Data lineage
You pay for two things: the compute consumed by automatic lineage reporting from supported Google Cloud systems (billed through the Dataplex premium processing SKU, in DCU-hours), and the storage of the resulting lineage metadata—processes, runs, lineage events, and links—for as long as it is retained (by default, 30 days; billed through the Knowledge Catalog metadata storage SKU, per GiB-month; see Metadata storage pricing). The billing appears in the project where you ran the job.
To separate data lineage charges from other charges in the Dataplex premium processing and metadata storage SKUs, use the label goog-dataplex-workload-type with the value LINEAGE on the Cloud Billing report.
Billable and non-billable usage
The following activities incur charges:
- Automatic lineage parsing: consumes Data Compute Unit (DCU) hours based on the processing complexity. DCUs measure the abstract compute capacity allocated to perform the parsing. Consumption is calculated based on the number of jobs, job executions, and the number of lineage links created.
- Lineage metadata storage: lineage metadata (processes, runs, lineage events, and links) is retained for a limited period (by default, 30 days) and billed under the metadata storage SKU for as long as it is stored. Storage charges stop when the metadata expires at the end of the retention period or when you delete it.
You incur DCU costs when the Data Lineage API Origin sourceType is set to any value other than CUSTOM. Values such as BIGQUERY, DATAPROC, or COMPOSER indicate that the lineage is being automatically extracted by Google Cloud from the service's logs and metadata. This extraction process requires compute resources (DCUs), which is why it is billable. In contrast, CUSTOM indicates that you have performed the parsing yourself, so no premium processing (DCU) charge applies; the stored result is subject to metadata storage pricing like any other lineage metadata.
The following activities are free:
- Custom lineage reporting: reporting lineage from third-party systems or custom integrations (where the sourceType is CUSTOM) incurs no premium processing charge. Metadata storage pricing still applies to the stored lineage. This includes the following methods:
- Calling ProcessOpenLineageRunEvent, which allows you to import OpenLineage-compatible events.
- Creating a LineageEvent using a custom client to manually report native lineage events.
- Reporting processes and runs: reporting the process metadata itself is free; charges only apply to the automatic parsing of LineageEvents.
- Lineage consumption: retrieving lineage (using Get or List methods) and deleting lineage (using Delete methods) are free.
Pricing example
In this example, you enable data lineage to track lineage for BigQuery in your project. The project is in the us-central1 location, where one DCU-hour costs $0.089.
Reporting lineage to or from one EntityReference costs approximately 0.0056 DCU-hours. An EntityReference uniquely identifies a data resource within Google Cloud, such as a BigQuery table or a Cloud Storage bucket.
- Daily job: You execute the following BigQuery job once a day for 30 days:
CREATE TABLE tableB1 AS SELECT * FROM tableA JOIN tableC USING (column);
Result: This operation generates lineage links connecting two source tables to one target table (3 distinct entities).
- Every other day job: You execute the following job every other day (15 times a month):
CREATE TABLE tableB2 AS SELECT * FROM tableA, tableC, tableD;
Result: This operation generates lineage links connecting three source tables to one target table (4 distinct entities).
Cost calculation
For this month, assume 1 GiB of data lineage metadata was generated. Costs are calculated as follows:
- Daily job: 30 days * 3 entities * ~0.0056 DCU-hours * $0.089 ~= $0.045
- Every other day job: 15 days * 4 entities * 0.0056 DCU-hours * $0.089 = $0.030
- Storage: (1 GiB) * $2.00 = $1.998
Total: ~$0.045+$0.030+$1.998 = ~$2.07
Design-time lineage pricing (Looker Core)
Lineage from Looker is design-time lineage: Knowledge Catalog derives it from the model definitions (for example, LookML views and Explores and the BigQuery tables they reference) rather than from job executions. Design-time lineage is billed under the same Dataplex premium processing SKU as automatic lineage parsing, but DCU-hour consumption is calculated differently:
- Charges apply the first time a lineage link or process is reported, and each time it changes (for example, when the Looker model is modified and re-published). There are no runs in design-time lineage, so there are no per-event charges, similar to event based lineage processing described above.
- Reporting lineage to or from one node costs approximately 0.0325 DCU-hours (approximately $0.0029 in us-central1, where one DCU-hour costs $0.089).
- Retrieving and deleting lineage remains free. Lineage metadata is charged separately under the metadata storage SKU.
Billing for Looker Core design-time lineage processing starts on March 1, 2027. Looker lineage reported before that date is free of charge. In order to opt out see https://docs.cloud.google.com/dataplex/docs/control-lineage-ingestion with information about enabling or disabling the Looker Core lineage processing.
