Fabric Spark is a cornerstone of Microsoft Fabric's unified analytics platform—blending lakehouse scale with natively integrated data engineering. Yet, unlocking consistent and optimized performance from Spark workloads in Fabric demands some understanding of the platform’s runtime, orchestration, and storage layers.

In this full-day technical workshop, Estera Kot and Amanjeet Singh will guide you through the core principles, tools, and architectural patterns for performance engineering with Spark in Microsoft Fabric. This is not generic Spark content—this session is built from real customer engagements, Fabric CAT insights, and platform-specific behaviors that directly impact your job performance, cost efficiency, and data reliability.

You will learn, hands-on and in-depth:
1 ) How Spark is Embedded in the Fabric Runtime
Understand how Spark jobs operate across notebooks, dataflows, and pipelines—and how Fabric’s execution engine, concurrency controls, and orchestration model affect throughput and performance.
2) Lakehouse & OneLake Data Optimization
Best practices for organizing data in OneLake using Delta Tables, V-Order, Z-Order, and Shortcut-based architecture—and how that structure impacts query performance, pipeline latency, and ingestion speed.
3 ) Performance Monitoring with Fabric-native Tools
Learn how to diagnose and optimize workloads using Monitoring Hub, Spark job insights, and Fabric Profiler—with actionable metrics like CU consumption, I/O bottlenecks, and job health status.
4 ) Designing Scalable Pipelines
We’ll walk through building efficient Data Factory pipelines using Spark activities—focusing on parallelism, resource reuse, and intelligent triggering for low-latency processing.

5 ) Capacity Units (CUs) and Cost-Aware Performance
How Fabric’s CU-based model differs from cluster-based provisioning, and what strategies help you balance speed and cost for Spark notebooks, jobs, and transformations.

6 ) Built-in Optimization Features
Explore how Native Execution Engine (NEE), query folding, and Copilot-assisted code generation can simplify and enhance Spark performance tuning.

7) Typical Anti-patterns in Fabric Spark
Real-world examples from enterprise use cases—common mistakes with storage access, job orchestration, or memory configurations—and how to fix them using platform-native techniques.

This workshop equips every data engineer and architect with not just patterns—but a framework to predict, measure, and optimize Spark jobs specifically inside Microsoft Fabric.