Phase 7 Telemetry Subsystem100% In-Browser Local

Fabric Capacity Observability & Diagnostics

Ingest official Microsoft Fabric Capacity Metrics App exports or test synthetic stress scenarios. Evaluate 24-hour background smoothing, pinpoint peak query spikes, and predict throttling before users are impacted.

Zero Exfiltration Guarantee: Your Fabric telemetry data is processed entirely in your browser memory. No operational logs, timestamps, or dataset names are ever transmitted to any external server or AI provider.

Active Telemetry Dataset

Enterprise Medallion Lakehouse (F64 Healthy)59 records (23h span)

Test Scenarios:

Evaluated Capacity:

Peak Instantaneous Demand

32.8 CUs

Recorded: 09:00

24h Smoothed Peak

46.2 CUs

Amortized across background window

Throttling Events

0incidents

Risk Index: 0 / 100

Interactive vs Background

59% / 41%

Interactive: 5m | Batch: 24h smooth

Recommended Safe Tier

F64 (64 CUs)

Includes +20% operating headroom

Reality Check: Active Scenario Alignment

Observed Telemetry vs Planned Scenario (Enterprise Production Baseline (F64 Target))

Optimal

Planned Capacity Model:

F128 (128 CUs)

Derived from active scenario inputs in Workspace.

Observed Smoothed Demand:

46.2 CUs (Peak: 32.8 CUs)

Utilization: 36% | Headroom: 64%

Engineering Guidance:

Maintain current F128 sizing. Monitor monthly growth to adjust before utilization exceeds 80%.

Hourly CU Consumption & 24-Hour Smoothing Timeline

Visualizes raw interactive and background CU demand along with Microsoft Fabric's 24-hour background smoothing curve against the F64 capacity threshold.

Background CU

Interactive CU

Smoothed Demand

Limit (64 CU)

Workload Engine Distribution

Proportion of total CU-hours consumed across Microsoft Fabric engines during the observed timeframe.

SemanticModel259.6 CU-h (58.8%)

Notebook60 CU-h (13.6%)

Pipeline50 CU-h (11.3%)

Eventhouse48 CU-h (10.9%)

Almacén24 CU-h (5.4%)

Smoothing & Throttling Mechanics

Microsoft Fabric Specification

Interactive Operations (5-Min Smoothing)

Power BI report renders, DAX queries, and lakehouse previews are smoothed over a 5-minute sliding window. Sudden concurrent spikes can immediately trigger Interactive Delays (UI pauses) if borrow thresholds are exceeded.

Background Operations (24-Hour Smoothing)

Spark notebooks, Data Factory pipelines, and Lakehouse compaction are smoothed over a 24-hour sliding window. A heavy batch job borrows capacity against future hours without immediate throttling, provided it clears within 24 hours.

Interactive Delay Threshold (10 Min)

When continuous consumption exceeds capacity limit for 10 consecutive minutes, subsequent user queries experience artificial delays to protect compute stability.

Cumulative Rejection Threshold (60 Min)

If accumulated borrow debt reaches the equivalent of 60 minutes of full capacity starvation, new interactive operations are rejected with HTTP 429 throttling errors.

Top Capacity-Consuming Artifacts

Identify which workspaces, notebooks, models, or pipelines are driving your monthly compute costs.

Filter by Engine:

Artifact / Item Name	Engine / Kind	Total CU-Hours	% of Capacity	Total Operations	Peak Burst CU
Executive Financial Model	SemanticModel	177 CU-h	40.1%	11	59 CU
Global Sales Dashboard	SemanticModel	82.6 CU-h	18.7%	11	28 CU
Silver Delta Cleansing	Notebook	60 CU-h	13.6%	5	38 CU
Raw Ingestion Pipeline	Pipeline	50 CU-h	11.3%	5	32 CU
IoT Fleet Telemetry Eventhouse	Eventhouse	48 CU-h	10.9%	24	4 CU
Gold Enterprise Warehouse	Warehouse	24 CU-h	5.4%	3	25 CU

Want to correlate this telemetry with your multi-region FinOps budget and architecture topology?Open Unified Engineering Workspace