The Mechanics of Fabric Capacity Unit (CU) Smoothing
Microsoft Fabric uses a dynamic smoothing algorithm to distribute interactive and background compute spikes across 24-hour time windows. Background jobs (Spark notebooks, Data Factory pipelines) are smoothed over 24 hours, while interactive queries (Power BI DirectLake visuals, SQL Warehouse endpoints) are smoothed over 5-minute to 1-hour windows. Understanding carryforward cumulative consumption is essential to prevent sudden throttling events.
Implementation & Verification Checklist
4 Verification Checks- Establish baseline CU consumption across interactive and background workloads
- Monitor carryforward CU burn-down rates during peak ETL processing windows
- Identify runaway DAX queries or inefficient Spark transformations causing persistent carryforward buildup
- Configure proactive alert thresholds at 70%, 85%, and 95% cumulative CU utilization
P-SKU to F-SKU Migration Economics & Auto-Pause ROI
Migrating from Power BI Premium (P-SKU) to Microsoft Fabric Capacity (F-SKU) introduces flexible Pay-As-You-Go (PAYG) and Reserved Instance economics. By automating capacity pauses during non-business hours (e.g. 12h/day auto-pause on Dev/Test capacities), organizations reduce compute spend by up to 50% compared to static 24/7 dedicated instances.
Implementation & Verification Checklist
4 Verification Checks- Calculate true TCO across PAYG, 1-Year Reserved, and 3-Year Reserved Fabric SKUs
- Verify F64+ SKU tiering to ensure free Power BI viewer licensing benefits apply to business users
- Implement automated pause and resume schedules for non-production environments via Azure Resource Manager APIs
- Evaluate Direct Lake memory capacity requirements against target F-SKU limits to prevent DirectQuery fallbacks
Two-Person Governed Action Gates for Capacity Changes
High-impact operational actions—such as capacity scaling, auto-pause schedule updates, or tenant workspace reallocations—require multi-eye authorization in enterprise environments. Enforcing cryptographic two-person approval gates prevents accidental service disruptions and satisfies strict SOC 2 and ISO 27001 change management compliance standards.
Implementation & Verification Checklist
4 Verification Checks- Require dual-authorization for any capacity scale-up, scale-down, or pause command
- Record tamper-evident SHA-256 cryptographic evidence hashes for all approved production actions
- Maintain an immutable 365-day audit trail of all requester and approver operational logs
- Integrate webhook notifications into Microsoft Teams and Slack for real-time sign-off workflows