Building high-performance analytics with Fabric Data Warehouse | BRK350
Rakesh Krishnan and Maraki Ketema present the architectural choices behind Fabric Data Warehouse and demonstrate GPU-accelerated query execution. The session connects those capabilities to the growing demands of analytics and AI agents.
Architecture and data access
The session introduces Microsoft Fabric as a platform that unifies data, analytics, and AI workloads, then examines Fabric Data Warehouse’s open, SaaS-first architecture.
Key areas include:
- Serverless, lake-centric design: How the warehouse architecture supports enterprise-scale analytics.
- OneLake integration: How the warehouse works with Fabric’s shared data foundation.
- Zero-copy data access: Accessing data without maintaining additional copies.
- Automatic scaling: Supporting changing workload demands alongside reliability and predictable performance.
Performance for analytics and AI workloads
Krishnan and Ketema discuss how AI agents drive higher query concurrency and create new demands on data warehouses. They cover GPU-accelerated Fabric Data Warehouse execution, including its performance and cost benefits, followed by a live demonstration of query acceleration.
The session also covers the SQL analytics endpoint, workload management, and built-in AI experiences, before moving to enterprise features, SQL enhancements, and migration tools.
Session chapters
- 00:00 — Introduction: Microsoft Fabric and Fabric Data Warehouse.
- 02:12 — AI-era challenges: The need for an AI-ready data platform.
- 04:38 — Unified workloads: How Fabric brings together data, analytics, and AI.
- 07:23 — Warehouse architecture: Serverless, lake-centric design.
- 09:29 — Scaling and access: Automatic scaling, OneLake integration, and zero-copy data access.
- 12:52 — AI-agent demands: Massive query concurrency and changing warehouse requirements.
- 15:20 — GPU acceleration: Performance and cost benefits.
- 23:12 — Live demonstration: Query acceleration and GPU-accelerated workloads.
- 28:01 — New capabilities: SQL analytics endpoint, workload management, and AI-native features.
- 36:15 — Enterprise capabilities: SQL enhancements, migration tools, and closing announcements.