top of page

Power BI Databricks Integration: Direct Lake, Unity Catalog and Query Performance

Writer: GrowthBI
GrowthBI
7 minutes ago
5 min read

Databricks and Power BI are a common pairing. Databricks handles data engineering and large-scale processing. Power BI gives business users the reports and dashboards they need. But how you connect the two has a big effect on speed, cost and security.


This guide explains the connection options, including Direct Lake through Microsoft Fabric, how Unity Catalog fits in and what to do when reports run slowly.


1. Connection options at a glance


Option

How it works

Best for

Import

Power BI copies data from Databricks into the model on refresh

Most reports, fastest visuals, data that refreshes on a schedule

DirectQuery

Power BI sends queries to a Databricks SQL warehouse on every interaction

Very large data or near real-time needs

Composite

Mix of Import and DirectQuery tables

Large fact tables with small imported dimensions

Direct Lake via Fabric

Power BI reads Delta tables in OneLake directly, using mirrored Unity Catalog data

Fast reports on large Delta tables without scheduled copies


2. Connecting Power BI to Databricks


The Azure Databricks connector


Power BI Desktop includes an Azure Databricks connector. You point it at a Databricks SQL warehouse using the server hostname and HTTP path, then sign in with Microsoft Entra ID. A SQL warehouse is recommended over an all-purpose cluster for BI workloads, as it is built for concurrent SQL queries.


Publish to Power BI from Databricks


Databricks also lets you publish Unity Catalog tables to the Power BI Service directly, creating a semantic model with the connection already configured. This is a quick way to get started.


Authentication


Use Microsoft Entra ID with single sign-on where possible. This means queries run as the user, so Unity Catalog permissions apply. Service principals work well for scheduled refresh of Import models.


3. Direct Lake with Databricks


What Direct Lake is


Direct Lake is a Power BI storage mode in Microsoft Fabric. It reads Delta Parquet files in OneLake directly into memory, giving performance close to Import without copying data on a schedule.


How Databricks data gets there


Fabric can mirror an Azure Databricks Unity Catalog. The mirrored catalog appears in Fabric, and the Delta tables are accessible through OneLake shortcuts without copying the data. Power BI semantic models can then use Direct Lake on those tables.


When it makes sense


  • You already use Fabric capacity

  • Tables are large and Import refresh is slow or expensive

  • You want near-current data without DirectQuery latency


Things to check


  • Direct Lake has guardrails on table size and row counts by capacity size

  • Some features fall back to DirectQuery when limits are exceeded

  • Security needs to be planned across both Databricks and Fabric


4. Unity Catalog and security


What Unity Catalog does


Unity Catalog is Databricks' central governance layer. It manages access to catalogs, schemas, tables, views and columns, and records lineage.


Making security flow through to Power BI


  • DirectQuery with SSO: each user's identity is passed to Databricks, so Unity Catalog rules apply automatically.

  • Import mode: data is copied using the refresh account's permissions. You must recreate access rules in Power BI with row-level security.

  • Direct Lake: plan access in Fabric, either with OneLake security or with row-level security in the semantic model.


Good practice


  • Use views or curated gold tables for Power BI, not raw tables

  • Grant Power BI service principals only the access they need

  • Document which security model applies to which report


5. Query performance


Slow reports are the most common complaint. Here is where to look.


On the Databricks side


  • Use a SQL warehouse sized for BI. Serverless SQL warehouses start quickly and scale with demand.

  • Model as a star schema. Facts and dimensions, not wide flat tables.

  • Optimise Delta tables. Use liquid clustering or Z-ordering on common filter columns.

  • Pre-aggregate. Build summary tables for common dashboard views.

  • Keep statistics current so the query optimiser makes good choices.


On the Power BI side


  • Prefer Import or Direct Lake for interactive dashboards.

  • Limit visuals per page. Each visual sends its own query in DirectQuery.

  • Use aggregations. Import a summary table and let Power BI fall back to DirectQuery only for detail.

  • Reduce columns. Only bring in the fields reports use.

  • Check with Performance Analyzer to find slow visuals and DAX.


Watch costs


DirectQuery dashboards with many users can keep a SQL warehouse busy all day. Import or Direct Lake usually costs less for heavily used reports.


6. A recommended pattern for mid-market teams


  1. Engineer data in Databricks into gold tables in Unity Catalog.

  2. Expose curated views for reporting.

  3. Use Import for most reports, refreshed with a service principal.

  4. Use DirectQuery or composite models for very large or time-sensitive tables.

  5. If on Fabric, mirror Unity Catalog and use Direct Lake for large tables.

  6. Apply row-level security in Power BI where data is imported.

  7. Monitor query time and warehouse cost monthly.


7. How GrowthBI helps


GrowthBI helps mid-market businesses with $30M to $500M in revenue connect Databricks and Power BI properly. We choose the right storage mode, align Unity Catalog and Power BI security and tune performance so reports load fast without runaway costs.


Pricing


  • Project pricing: $5,000 to $30,000 AUD per project, depending on scope and data volumes.

  • Ongoing support and optimisation: from $2,000 per month.


Timeline


  • Most integration and performance projects take 1 to 6 weeks, depending on how much data preparation is needed.


Learn more about our data modernisation services.


Frequently Asked Questions


Can Power BI connect to Databricks?


Yes. Power BI connects to Databricks SQL warehouses through the Azure Databricks connector using Import, DirectQuery or composite models. Databricks can also publish Unity Catalog tables directly to the Power BI Service.


What is Direct Lake?


Direct Lake is a Power BI storage mode in Microsoft Fabric that reads Delta tables in OneLake directly. With Databricks, it works through Fabric mirroring of Unity Catalog.


Does Unity Catalog security apply in Power BI?


It applies when users query through DirectQuery with single sign-on. In Import mode, data is copied using the refresh account, so you need row-level security in Power BI.


Should I use Import or DirectQuery with Databricks?


Import is usually faster and cheaper for interactive dashboards. DirectQuery suits very large datasets or near real-time needs. Composite models can combine both.


How do I speed up Power BI reports on Databricks?


Use a BI-sized SQL warehouse, a star schema, optimised Delta tables, pre-aggregated tables, fewer visuals per page and Power BI aggregations.


Make Databricks and Power BI Work as One


The right connection pattern gives users fast reports, keeps security consistent and controls cost. Get it right once and every new report benefits.


Book a free consultation with GrowthBI to review your Databricks and Power BI setup.

 
 

Recent Posts

See All
bottom of page