Databricks
Every Databricks DBU, traced to a job.
Cloud Intelligence™ imports your Databricks cost and usage from system tables: DBUs by workspace, job, cluster, SQL warehouse and serving endpoint, plus cluster CPU and memory. Right next to the rest of your cloud spend.
Service principal · 6 months of history · Tracked hourly
Databricks · October
what we import
Databricks data, in the dimensions you already use.
Every charge maps to standard Cloud Analytics dimensions, plus Databricks labels for cluster, warehouse, job, notebook, pipeline, endpoint, run_as identity and features like serverless and Photon.
acme-databricks1234567890123456prod-etlSQLENTERPRISE_ALL_PURPOSE_COMPUTE_(PHOTON)DBUORIGINALwhat you can do with it
From the Databricks bill to a decision
Cost by job and cluster
Slice by workspace, job, cluster, notebook, pipeline or serving endpoint.
Serverless vs classic
See how much of your DBU spend runs on serverless compute.
Jobs vs all-purpose
DBU usage and cost by workload type, plus cluster CPU and memory.
Databricks Intelligence
A preset dashboard, plus Databricks next to your other providers.
Budgets and alerts
Guardrails on DBU spend, plus alerts when we spot cost anomalies.
Allocations
Put Databricks spend on teams by workspace, job or identity.
live in hours
Connect Databricks in three steps
Create a service principal.
Generate an OAuth secret and grant it read access to billing and compute system tables.
Connect in the catalog.
Enter your account name, client ID, secret and a SQL warehouse.
Reports in hours.
Up to 6 months of history imported. We email you when it's ready.
Read-only access to system tables. Disconnect any time.
Start freeFrequently asked
questions
How do we connect Databricks?
Open the Databricks card in the integrations catalog and follow the wizard: enter your account name, the client ID and secret of a service principal, and at least one SQL warehouse ID and server hostname. Test the connection, then finish.
What permissions are needed?
You need to be a Databricks Account Admin and metastore admin, with Unity Catalog enabled. The service principal gets SELECT on system.billing, system.compute and system.access tables, plus the Databricks SQL access entitlement.
Why do we need a SQL warehouse?
We use it as a compute endpoint to query your system tables multiple times a day, even if you don't run Databricks SQL yourself. An existing warehouse works, or create a small classic one (2X-small, 5-minute auto stop).
How much history do we get, and how fresh is it?
We import the last 6 months of billing and usage data on connect, which takes a few hours. Data is then tracked and reported hourly.
Which dimensions and labels are available?
Account, workspace, product, SKU, usage unit and record type map to standard Cloud Analytics dimensions, so Databricks sits next to your other providers. Labels add cluster, warehouse, job, notebook, pipeline, endpoint, run_as identity and product features, and cluster CPU and memory utilization appear as extended metrics.
Do we need to connect every workspace?
No. Databricks billing data is account-level, so one warehouse covers all workspaces. Add a warehouse in each region only if you want fuller cluster utilization reporting.
Connect Databricks. Know what every DBU costs.
Free to start. Read-only. Reports in hours.