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:

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.

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:

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

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: Serverless usage from fine-grained access control on dedicated compute
• BASE_ENVIRONMENTS: Usage associated with building or refreshing a workspace's serverless base environment

• 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
• DATABASE: Costs associated with Lakebase database instances

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 is true or false, or null (value is true or false when you can choose between serverless and classic compute, otherwise it's 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.
• performance_target: Indicates the performance mode of the serverless job or pipeline. Values include PERFORMANCE_OPTIMIZED, STANDARD, or null. Non-serverless workloads have a null value.

• networking.connectivity_type: values include PUBLIC_IP and PRIVATE_IP