Fabric 용량 단위: 컴퓨팅, 라이선스, 가격 구분
CU의 의미와 워크로드 비교 방법을 이해하고 SKU를 사용자 수 보장으로 간주하지 마세요.
영어 참고 문서
What the number tells you
A Fabric capacity supplies a pool of compute resources for assigned workloads. F-SKU names identify capacity-unit sizes: F2 has two CUs and F64 has 64. That ratio does not promise that every query runs 32 times faster or that a fixed number of users will fit. Workload shape, concurrent operations, data models, and the service's resource rules influence the outcome.
A small budgeting exercise
Consider two hypothetical teams that both choose the same SKU. One refreshes a compact model overnight and serves a few morning queries. The other launches overlapping transformations while many readers open complex reports. Their capacity labels match, but their contention patterns can differ. Record a time-based workload schedule rather than adding a user count to a storage volume and calling the result a measured CU requirement.
Keep entitlements separate
Power BI licensing and compute sizing answer different questions. Organizational free-license viewers need appropriately assigned content and access on F64 or higher capacity. Paid F2 or higher can be eligible for Copilot, subject to tenant, region, and feature requirements. A larger capacity alone does not grant every permission or satisfy every licensing condition.
Verify before scaling
Collect operation-level evidence with the Capacity Metrics app and inspect the periods when user-facing behavior fails. Compare a baseline period with a representative peak, including refreshes and background work. Check whether a model or query change addresses the bottleneck before simply increasing the tier. Treat prices shown in this app as reference assumptions and verify the actual region, currency, and offer. A useful decision record explains which observed constraint the selected capacity is expected to address and how the team will test that expectation.