Performance
9 posts on Performance: benchmarks, internals and workarounds for the Microsoft data stack.
- 09How Much Does Direct Lake Cost in Microsoft Fabric?
Direct Lake vs Import CU cost, measured over nine replications on F64 and F2. Where the refresh saving goes, and the Lakehouse overhead you don't see.
- 08Direct Lake Internals: Transcoding, Framing, Residency, and Eviction
How Direct Lake really works in Microsoft Fabric: framing, transcoding, column residency and eviction, with the measured CU cost of each.
- 07Fabric Dataflow Gen2 Partitioned Compute: Setup and Benchmark
Discover how to set up Partitioned Compute in Dataflow Gen2 for parallel processing. Read my setup guide and see the real-world benchmark results in Fabric.
- 06Automated Delta Table Maintenance in Fabric Lakehouse (Without PySpark)
Deep dive into Delta table internals, transaction logs, and storage layout. Master maintenance, VACUUM, and Z-Order using Python (delta-rs) in Microsoft Fabric.
- 05Fix Slow Refreshes: Faster Way to Load SharePoint Files in Power BI
Optimize Power BI SharePoint data retrieval. Learn technical differences between connectors and implement a custom REST API strategy for faster dataset refreshes.
- 04How to Fix Slow SharePoint List Refresh in Power BI
Boost Power BI SharePoint refresh speed. Fix RenderListDataAsStream errors and use a custom M script to cut load times from minutes to seconds.
- 03Why Is Power Query List.Contains Slow? Faster Lookup Alternatives
Power Query lookup benchmarks. Learn why List.Contains is slow at scale, when buffering helps, and how joins or Record.FieldOrDefault beat list contains.
- 02How Power Query (M) evaluates your queries, and why the performance gets weird
Lazy versus eager evaluation in Power Query M, what buffering actually changes, and why the same query can return different numbers on each run.
- 01Stop Repeating CALCULATE: Iterate Over Categories in DAX
Stop repeating CALCULATE in your DAX. Discover how to iterate and simplify category-level logic with SUMX, AVERAGEX, and virtual tables like VALUES and SUMMARIZE.
