balajij8
Esteemed Contributor II

‎06-20-2026 09:47 AM - edited ‎06-20-2026 10:05 AM

You can follow below

Serverless Optimization

  • Implement Delta Lake optimizations (liquid clustering, Z-ordering) on the tables to reduce scan costs and improve query performance
  • Use query result caching - Power BI can reuse cached results for identical queries
  • Create materialized views for reporting instead of querying raw data repeatedly. Use partition pruning and predicate pushdown

Refresh Approach

  • Validate if real-time is required or if scheduled incremental refreshes (every 30 minutes) meet needs
  • Use Power BI's incremental refresh feature to only refresh new/changed data & not full datasets

Alternatives

Provisioned SQL Warehouses

  • Classic SQL Warehouses with auto-stop can be more cost-effective for predictable workloads
  • For intermittent usage, configure auto stop (10 minutes)
  • For continuous usage, configure a right-sized always on warehouse as it generally costs less than serverless

Hybrid

  • Use Serverless for ad-hoc queries and exploration
  • Use provisioned warehouses for scheduled Power BI refreshes with predictable compute needs
 

Databricks Streaming & Direct Lake

You can evaluate below if using Azure Databricks

  • You can expose curated Delta from Azure Databricks to Fabric One Lake and leverage Direct Lake mode for Power BI reporting. Power BI queries the data directly from the storage for report rendering. Fabric costs will be added to the bill with additional considerations.

Power BI Mode

  • Consider Import mode with scheduled refresh instead of Direct Query if real time isn't required
  • Consider Composite models - Import aggregated data, Direct Query for details only when needed if possible
0 Kudos