Billable usage system table reference

This article provides an overview of the billable usage system table, including the schema and example queries. With system tables, your account's billable usage data is centralized and routed to all regions, so you can view your account's global usage from whichever region your workspace is in.

For information on using this table to monitor costs and sample queries, see Monitor costs using system tables.

Table path: This system table is located at system.billing.usage.

Billable usage table schema

The billable usage system table uses the following schema:

Column name	Data type	Description	Example
record_id	string	Unique ID for this usage record	11e22ba4-87b9-4cc2-9770-d10b894b7118
account_id	string	ID of the account this report was generated for	23e22ba4-87b9-4cc2-9770-d10b894b7118
workspace_id	string	ID of the workspace this usage was associated with	1234567890123456
sku_name	string	Name of the SKU	STANDARD_ALL_PURPOSE_COMPUTE
cloud	string	Cloud associated with this usage. Possible values are AWS, AZURE, and GCP.	AWS, AZURE, or GCP
usage_start_time	timestamp	The start time relevant to this usage record. Timezone information is recorded at the end of the value with +00:00 representing UTC timezone.	2023-01-09 10:00:00.000+00:00
usage_end_time	timestamp	The end time relevant to this usage record. Timezone information is recorded at the end of the value with +00:00 representing UTC timezone.	2023-01-09 11:00:00.000+00:00
usage_date	date	Date of the usage record, this field can be used for faster aggregation by date	2023-01-01
custom_tags	map	Custom tags associated with the usage record	{ “env”: “production” }
usage_unit	string	Unit this usage is measured in	DBU
usage_quantity	decimal	Number of units consumed for this record	259.2958
usage_metadata	struct	System-provided metadata about the usage, including IDs for compute resources and jobs (if applicable). See Usage Metadata.	See Usage metadata
identity_metadata	struct	System-provided metadata about the identities involved in the usage. See Identity Metadata.	See Identity metadata
record_type	string	Whether the record is original, a retraction, or a restatement. The value is ORIGINAL unless the record is related to a correction. See Record Type.	ORIGINAL
ingestion_date	date	Date the record was ingested into the usage table	2024-01-01
billing_origin_product	string	The product that originated the usage. Some products can be billed as different SKUs. For possible values, see Product.	JOBS
product_features	struct	Details about the specific product features used. See Product features.	See Product features
usage_type	string	The type of usage attributed to the product or workload for billing purposes. Possible values are COMPUTE_TIME, STORAGE_SPACE, NETWORK_BYTES, NETWORK_HOUR, API_OPERATION, TOKEN, or GPU_TIME.	STORAGE_SPACE

Usage metadata reference

The values in usage_metadata are all strings that tell you about the workspace objects and resources involved in the usage record.

Only a subset of these values is populated in any given usage record, depending on the compute type and features used. The third column in the table shows which usage types cause each value to be populated.

Value	Description	Populated for (otherwise null)
cluster_id	ID of the cluster associated with the usage record	Non-serverless compute usage, including notebooks, jobs, DLT, and legacy model serving
job_id	ID of the job associated with the usage record	Serverless jobs and jobs run on job compute (does not populate for jobs run on all-purpose compute)
warehouse_id	ID of the SQL warehouse associated with the usage record	Workloads run on a SQL warehouse
instance_pool_id	ID of the instance pool associated with the usage record	Non-serverless compute usage from pools, including notebooks, jobs, DLT, and legacy model serving
node_type	The instance type of the compute resource	Non-serverless compute usage, including notebooks, jobs, DLT, and all SQL warehouses
job_run_id	ID of the job run associated with the usage record	Serverless jobs and jobs run on job compute (does not populate for jobs run on all-purpose compute)
notebook_id	ID of the notebook associated with the usage	Serverless notebooks
dlt_pipeline_id	ID of the DLT pipeline associated with the usage record	DLT and features that use DLT pipelines, such as materialized views, online tables, vector search indexing, and Lakeflow Connect
endpoint_name	The name of the model serving endpoint or vector search endpoint associated with the usage record	Model serving and Vector Search
endpoint_id	ID of the model serving endpoint or vector search endpoint associated with the usage record	Model serving and Vector Search
dlt_update_id	ID of the DLT pipeline update associated with the usage record	DLT and features that use DLT pipelines, such as materialized views, online tables, vector search indexing, and Lakeflow Connect
dlt_maintenance_id	ID of the DLT pipeline maintenance tasks associated with the usage record	DLT and features that use DLT pipelines, such as materialized views, online tables, vector search indexing, and Lakeflow Connect
metastore_id	This value is not populated in Azure Databricks	Always null
run_name	Unique user-facing name of the Foundation Model Fine-tuning run associated with the usage record	Foundation Model Fine-tuning
job_name	User-given name of the job associated with the usage record	Jobs run on serverless compute
notebook_path	Workspace storage path of the notebook associated with the usage	Notebooks run on serverless compute
central_clean_room_id	ID of the central clean room associated with the usage record	Clean Rooms
source_region	Region of the workspace associated with the usage. Only returns a value for serverless networking-related usage.	Serverless networking
destination_region	Region of the resource being accessed. Only returns a value for serverless networking-related usage.	Serverless networking
app_id	ID of the app associated with the usage record	Databricks Apps
app_name	User-given name of the app associated with the usage record	Databricks Apps
private_endpoint_name	Name of the applicable private endpoint deployed with serverless compute	Serverless networking
budget_policy_id	ID of the serverless budget policy attached to the workload	Serverless compute usage, including notebooks, jobs, DLT, and model serving endpoints

Identity metadata reference

The identity_metadata column provides more information about the identities involved in the usage.

• The run_as field logs who ran the workload. This values is only populated for certain workload types listed in the table below.
• The owned_by field only applies to SQL warehouse usage and logs the user or service principal who owns the SQL warehouse responsible for the usage.

• The identity_metadata.created_by field applies to Databricks Apps and logs the email of the user who created the app.

run_as identities

The identity recorded in identity_metadata.run_as depends on the product associated with the usage. Reference the following table for the identity_metadata.run_as behavior:

Workload type	Identity of run_as
Jobs compute	The user or service principal defined in the run_as setting. By default, jobs run as the identity of the job owner, but admins can change this to be another user or service principal.
Serverless compute for jobs	The user or service principal defined in the run_as setting. By default, jobs run as the identity of the job owner, but admins can change this to be another user or service principal.
Serverless compute for notebooks	The user who ran the notebook commands (specifically, the user who created the notebook session). For shared notebooks, this includes usage by other users sharing the same notebook session.
DLT pipelines	The user whose permissions are used to run the DLT pipeline. This can be changed by transferring the pipeline's ownership.
Foundation Model Fine-tuning	The user or service principal that initiated the fine-tuning training run.
Predictive optimization	The Databricks-owned service principal that runs predictive optimization operations.
Lakehouse monitoring	The user who created the monitor.

Record type reference

The billing.usage table supports corrections. Corrections occur when any field of the usage record is incorrect and must be fixed.

When a correction happens, Azure Databricks adds two new records to the table. A retraction record negates the original incorrect record, then a restatement record includes the corrected information. Correction records are identified using the record_type field:

• RETRACTION: Used to negate the original incorrect usage. All fields are identical to the ORIGINAL record except usage_quantity, which is a negative value that cancels out the original usage quantity. For example, if the original record's usage quantity was 259.4356, then the retraction record would have a usage quantity of -259.4356.
• RESTATEMENT: The record that includes the correct fields and usage quantity.

For example, the following query returns the correct hourly usage quantity related to a job_id, even if corrections have been made. By aggregating the usage quantity, the retraction record negates the original record and only the restatement's values are returned.

SELECT
  usage_metadata.job_id, usage_start_time, usage_end_time,
  SUM(usage_quantity) as usage_quantity
FROM system.billing.usage
GROUP BY ALL
HAVING usage_quantity != 0

Note

For corrections where the original usage record should not have been written, a correction may only add a retraction record and no restatement record.

Billing origin product reference

Some Databricks products are billed under the same shared SKU. For example, Lakehouse Monitoring, predictive optimization, and serverless workflows are all billed under the same serverless jobs SKU.

To help you differentiate usage, the billing_origin_product and product_features columns provide more insight into the specific product and features associated with the usage.

The billing_origin_product column shows the Databricks product associated with the usage record. The values include:

• JOBS
• DLT
• SQL
• ALL_PURPOSE
• MODEL_SERVING
• INTERACTIVE
• DEFAULT_STORAGE
• VECTOR_SEARCH
• LAKEHOUSE_MONITORING
• PREDICTIVE_OPTIMIZATION
• ONLINE_TABLES
• FOUNDATION_MODEL_TRAINING
• AGENT_EVALUATION
• FINE_GRAINED_ACCESS_CONTROL

• NETWORKING: Costs associated with connecting serverless compute to your resources through private endpoints. For NETWORKING usage, workspace_id is null, usage_unit is hour, and networking.connectivity_type is PRIVATE_IP.
• APPS: Costs associated with building and running Databricks Apps

Product features reference

The product_features column is an object containing information about the specific product features used and includes the following key/value pairs:

• jobs_tier: values include LIGHT, CLASSIC, or null
• sql_tier: values include CLASSIC, PRO, or null
• dlt_tier: values include CORE, PRO, ADVANCED, or null
• is_serverless: values include true or false, or null
• is_photon: values include true or false, or null
• serving_type: values include MODEL, GPU_MODEL, FOUNDATION_MODEL, FEATURE, or null
• offering_type: values include BATCH_INFERENCE or null.

• networking.connectivity_type: values include PUBLIC_IP and PRIVATE_IP