Analyze billable usage log data

This article explains how to read and analyze the usage data from your Databricks account.

You can view and download billable usage directly in the account console.

CSV file schema

Column	Type	Description	Example
workspaceId	string	ID of the workspace.	1234567890123456
timestamp	datetime	End of the hour for the provided usage.	2019-02-22T09:59:59.999Z
clusterId	string	ID of the cluster (for a cluster) or of the warehouse (for a SQL warehouse)	Cluster example: 0406-020048-brawl507 SQL warehouse example: 8e00f0c8b392983e
clusterName	string	User-provided name for the cluster/warehouse.	Shared Autoscaling
clusterNodeType	string	Instance type of the cluster/warehouse.	Cluster example: m4.16xlarge SQL warehouse example: db.xlarge
clusterOwnerUserId	string	ID of the user who created the cluster/warehouse.	12345678901234
clusterCustomTags	string (“-escaped json)	Custom tags associated with the cluster/warehouse during this hour.	"{""dept"":""mktg"",""op_phase"":""dev""}"
sku	string	Billing SKU. See the Billing SKUs table for a list of values.	STANDARD_ALL_PURPOSE_COMPUTE
dbus	double	Number of DBUs used by the user during this hour.	1.2345
machineHours	double	Total number of machine hours used by all containers in the cluster/warehouse.	12.345
clusterOwnerUserName	string	Username (email) of the user who created the cluster/warehouse.	user@yourcompany.com
tags	string (“-escaped json)	Default and custom cluster/warehouse tags, and default and custom instance pool tags (if applicable) associated with the cluster during this hour. See Cluster tags, Warehouse tags, and Pool tags. This is a superset of the clusterCustomTags column.	"{""dept"":""mktg"",""op_phase"":""dev"", ""Vendor"":""Databricks"", ""ClusterId"":""0405-020048-brawl507"", ""Creator"":""user@yourcompany.com""}"

Billing SKUs

• ENTERPRISE_ALL_PURPOSE_COMPUTE

• ENTERPRISE_ALL_PURPOSE_COMPUTE_(PHOTON)

• ENTERPRISE_JOBS_COMPUTE

• ENTERPRISE_JOBS_COMPUTE_(PHOTON)

• ENTERPRISE_JOBS_LIGHT_COMPUTE

• ENTERPRISE_SQL_COMPUTE

• ENTERPRISE_DLT_CORE_COMPUTE

• ENTERPRISE_DLT_CORE_COMPUTE_(PHOTON)

• ENTERPRISE_DLT_PRO_COMPUTE

• ENTERPRISE_DLT_PRO_COMPUTE_(PHOTON)

• ENTERPRISE_DLT_ADVANCED_COMPUTE

• ENTERPRISE_DLT_ADVANCED_COMPUTE_(PHOTON)

• PREMIUM_ALL_PURPOSE_COMPUTE

• PREMIUM_ALL_PURPOSE_COMPUTE_(PHOTON)

• PREMIUM_JOBS_COMPUTE

• PREMIUM_JOBS_COMPUTE_(PHOTON)

• PREMIUM_JOBS_LIGHT_COMPUTE

• PREMIUM_SQL_COMPUTE

• PREMIUM_DLT_CORE_COMPUTE

• PREMIUM_DLT_CORE_COMPUTE_(PHOTON)

• PREMIUM_DLT_PRO_COMPUTE

• PREMIUM_DLT_PRO_COMPUTE_(PHOTON)

• PREMIUM_DLT_ADVANCED_COMPUTE

• PREMIUM_DLT_ADVANCED_COMPUTE_(PHOTON)

• STANDARD_ALL_PURPOSE_COMPUTE

• STANDARD_ALL_PURPOSE_COMPUTE_(PHOTON)

• STANDARD_JOBS_COMPUTE

• STANDARD_JOBS_COMPUTE_(PHOTON)

• STANDARD_JOBS_LIGHT_COMPUTE

• STANDARD_DLT_CORE_COMPUTE

• STANDARD_DLT_CORE_COMPUTE_(PHOTON)

• STANDARD_DLT_PRO_COMPUTE

• STANDARD_DLT_PRO_COMPUTE_(PHOTON)

• STANDARD_DLT_ADVANCED_COMPUTE

• STANDARD_DLT_ADVANCED_COMPUTE_(PHOTON)

Analyze usage data in Databricks

This section describes how to make the data in the billable usage CSV file available to Databricks for analysis. It describes options for creating a usage table and includes a sample notebook that you can use to run a usage analysis dashboard.

The CSV file uses a format that is standard for commercial spreadsheet applications but requires a modification to be read by Apache Spark. You must use option("escape", "\"") when you create the usage table in Databricks.

Total DBUs are the sum of the dbus column.

Import the log using the Create Table UI

You can use the Load data using the add data UI to import the CSV file into Databricks for analysis.

Create a Spark DataFrame

You can also use the following code to create the usage table from a path to the CSV file:

df = (spark.
      read.
      option("header", "true").
      option("inferSchema", "true").
      option("escape", "\"").
      csv("/FileStore/tables/usage_data.csv"))

df.createOrReplaceTempView("usage")

If the file is stored in an S3 bucket, for example when it is used with log delivery, the code will look like the following. You can specify a file path or a directory. If you pass a directory, all files are imported. The following example specifies a file.

df = (spark.
      read.
      option("header", "true").
      option("inferSchema", "true").
      option("escape", "\"").
      load("s3://<bucketname>/<pathprefix>/billable-usage/csv/workspaceId=<workspace-id>-usageMonth=<month>.csv"))

df.createOrReplaceTempView("usage")

The following example imports a directory of billable usage files:

df = (spark.
      read.
      option("header", "true").
      option("inferSchema", "true").
      option("escape", "\"").
      load("s3://<bucketname>/<pathprefix>/billable-usage/csv/"))

df.createOrReplaceTempView("usage")

Create a Delta table

To create a Delta table from the DataFrame (df) in the previous example, use the following code:

(df.write
    .format("delta")
    .mode("overwrite")
    .saveAsTable("database_name.table_name")
)

Warning

The saved Delta table is not updated automatically when you add or replace new CSV files. If you need the latest data, re-run these commands before you use the Delta table.

Usage analysis dashboard notebook

If you use billable usage delivery, you can use the following notebook to run a usage analysis dashboard by providing a path to the S3 bucket where your CSV files are stored and entering report parameters in a widget.

The widget that you use to enter report parameters appears above the first notebook cell when you import the notebook to your Databricks workspace. The widget does not appear in the browser-only view of the notebook. Here’s an image of the widget:

Usage analysis dashboard notebook

Open notebook in new tab Copy link for import