Fabric Weekly
Issue 4
28th July 2026
🎉 Welcome to issue #4 of Fabric Weekly!
On the official side:
- Native Execution Engine now accelerating CSV workloads in Fabric Spark - up to 2x faster, and no code changes needed if it's already enabled.
- Fabric data agents in Microsoft Foundry get easier discovery and trust via MCP, OneLake Catalog, and multi-agent support.
- And, Forrester has named Microsoft a Leader in its Multimodel Data Platforms Wave.
From the community:
- Turning on Fabric's inbound network protection sounds like an easy win - Meagan Longoria explains why you might want to think twice!
- Vojtěch Šíma provides a deep dive on Direct Lake internals.
- Is Azure Data Factory going away? Aleksi Partanen put the question straight to Microsoft PM Mark Kromer. Short answer: no, but the new stuff is landing in Fabric Data Factory first.
- And, James Serra refreshed his guide to Power BI performance in 2026.
🤖 Copilot, AI & Agents
- Fabric data agents in Microsoft Foundry: Easier to connect, easier to trust Fabric data agents in Microsoft Foundry are now easier to discover, integrate, and monitor with Model Context Protocol integration, OneLake Catalog discovery, multi-agent support, and improved logging.
- Prepare your Fabric Data Agent integrations for Assistants API retirement Migrating to Fabric data agent’s Model Context Protocol endpoint by August 26, 2026, is required as the OpenAI Assistants API will retire, preserving integration functionality.
- Microsoft recognized as a Leader in The Forrester Wave™: Multimodel Data Platforms Forrester recognizes Microsoft Fabric's ability to unify multimodel data governance, security, and AI integration, meeting the demand for efficient cohesive platforms.
📊 Reporting & Insights
- Simplify Power BI DAX with Visual Calculations Visual calculations in Power BI let you write intuitive functions like RUNNINGSUM([Sales]) directly within visuals, often speeding up performance but requiring careful consideration to avoid fragmenting business logic.
- Build event-driven architectures in Fabric with Real-Time Hub Fabric’s Real-Time Hub simplifies building event-driven architectures by integrating three core event pillars that let applications instantly respond to critical business events.
- How to Export Jira Data to Power BI: A Step-by-Step Guide for Every Team This guide outlines four methods for exporting Jira data to Power BI, including CSV exports, Atlassian connectors, and the Power BI Web API, enabling seamless integration of Jira insights into real-time dashboards for stakeholders.
- Power BI, M365 Copilot And The Importance Of DAX UDFs Adding DAX UDFs to Power BI models improves M365 Copilot's query accuracy and optimization while reducing performance risks and inconsistent results.
- Implementing Dynamic Bin Sizing with Multiple User Inputs in Power BI The video demonstrates how Power BI enables dynamic bin sizing based on several user-provided inputs, streamlining data analysis workflows.
- Power BI Performance in 2026: A Practical Guide to Faster Semantic Models and Reports This updated guide explains that Power BI performance results from multiple factors including data source choice, storage mode, model design, and refresh processes, emphasizing the need to diagnose the entire architecture for bottleneck resolution.
- Instantly Improve Slicers in Power BI Dashboards Discover five practical Power BI slicer techniques that enhance dashboard interactivity and usability.
⚙️ Data Engineering
- Is Azure Data Factory Going Away? I Asked the Microsoft PM Mark Kromer explains ADF will continue to be supported but new features like Copilot and agents will launch first in Fabric Data Factory, urging users to migrate for improved capabilities.
- Native Execution Engine now accelerates CSV workloads in Microsoft Fabric Spark Fabric Spark's Native Execution Engine now has a Velox-based, SIMD-optimized CSV reader that avoids row-to-columnar conversion overhead — up to 2x faster on TPC-DS CSV benchmarks, with smaller gains (~20-65%) on other workloads, and no code changes required.
- 95% of You Will Use This Fabric Feature Wrong Fabric data pipelines' new Approval activity (preview) has a few genuine use cases, but most people reaching for it — like manual per-load approvals every morning — are misusing their pipeline logic.
- Managing State in Data Pipelines Is a Pain. Spark Fixes That. Spark structured streaming's built-in state management removes the log-read-process-write-back round trip that data engineers otherwise build by hand for stateful pipelines.
- Intelligent streaming: Turning customer intent into revenue with Fabric Eventstreams, Spark, and AI Functions Fabric Eventstreams, Spark, and AI Functions analyze cart events in real time to offer personalized interventions, potentially recovering up to 12% of abandoned online shopping carts by responding instantly upon detecting intent.
- Spark in Fabric Is About to Get Insanely Fast Spark 4.1's upcoming real-time mode for Kafka and Event Hub endpoints brings double-digit-millisecond stream processing latency to Fabric Spark, down from about a second in micro-batch mode.
- Explore, validate, and iterate faster with the new Lakehouse Query Explorer in Microsoft Fabric (Generally Available) Lakehouse Query Explorer in Microsoft Fabric enables data professionals to run and iterate Spark SQL queries directly within the interface, enhancing validation and exploration speed.
🏛️ Storage & Platform
- Microsoft Admits: Non-Schema Lakehouses Have No Major Advantage Ted from Microsoft explains that non-schema lakehouses offer no significant advantage over schema-based ones, suggesting Fabric should adopt a single lakehouse type moving forward.
- How to Design a Bronze Layer in Microsoft Fabric That Actually Scales Designing a scalable bronze layer in Microsoft Fabric requires selecting storage options, architecture type, and implementing metadata-driven pipelines with JSON configurations.
- Direct Lake Internals: Transcoding, Framing, Residency, and Eviction Direct Lake on OneLake outlines its internal processes such as Delta-table framing, Parquet handling, VertiPaq transcoding, segment residency and eviction policies, prewarming strategies, and compares CU consumption to traditional Import methods.
- Lakehouse Schema Migration Is Nearly Instant (Any Table Size) Ted from Microsoft demonstrates that schema migrations for lakehouse environments occur instantly, without requiring a new lakehouse or affecting performance even on very large tables.
🔐 Governance & Security
- Think Twice Before Enabling Fabric’s Inbound Network Protection Enabling Fabric’s inbound network protection may degrade or disable critical features such as on-premises data gateways, mirroring for older SQL Server versions, and certain pipelines without clear documentation.
- Simplify OneLake network security for Azure Services with resource instance rules (Generally Available) Resource instance rules in OneLake let workspace admins securely grant service-to-service access from Azure services like Databricks or SQL Server without complex IP allowlists, using verified Azure resource identities to enforce fine-grained, layered security.
💰 Management & Cost
- My First Prompt: Build Me Microsoft Dynamics An experiment prompting Fabric Apps to generate a Microsoft Dynamics clone from a single instruction, with a token counter added to track the cost of the build.
- Five Years of Bringing DataOps to Power BI and Fabric Over five years, DataOps has expanded from Power BI to Fabric, reducing cycle times with features like Project files and User-Defined Functions, and improving reusability and quality via automated testing.