Pillar B
How to Safely Automate Fabric Capacity Pause/Resume
Automate Fabric capacity pause/resume safely: clear smoothed background debt first, then use the REST suspend/resume API. Safe checklist + dollar savings.
Jonathan Flach
Founder, SpendWeave
Founder of SpendWeave. Built SpendWeave's client-side Microsoft Fabric capacity-audit engine and writes about Fabric FinOps, capacity cost optimization, and the platform's billing mechanics from hands-on analysis of real capacity telemetry.
Articles by Jonathan Flach (34)
Pillar C
Alternatives to the Capacity Metrics App for Long-Term History & Attribution
The Capacity Metrics app keeps only 14 days of compute history. An honest build-vs-buy of the alternatives: SemPy vault, FUAM, and SpendWeave Pro.
Pillar B
Why Dataflows Gen2 Quietly Drain Your Fabric Capacity
Dataflows Gen2 can draw from four simultaneous billing meters that stack fast. See when notebooks or stored procedures cost less — with the math.
Pillar A
How to Estimate Your Monthly Microsoft Fabric Bill
Three-line formula to estimate your Fabric bill: CUs × $0.18 × 730, OneLake $0.023/GB-mo, PAYG overage. Includes a complete worked monthly-bill model.
Pillar C
Extract Fabric Capacity Metrics with SemPy: What the Data Reveals at Scale
SemPy's evaluate_dax bypasses the executeQueries 100k-row cap to pull the full Capacity Metrics model — and reveals the gap no script closes.
Pillar C
Fabric Cost Attribution: Why You Can't Identify Which Pipeline Run Cost You Money
Fabric's OperationID isn't linked to pipeline runs, attribution is item-level only, and the Chargeback app aggregates daily. Here's how to close the gap.
Pillar C
Fabric Capacity Events in the Real-Time Hub: Near-Real-Time Monitoring
Fabric capacity events fire every 30 s and on state change. Route via Eventstream to an Eventhouse + Data Activator for sub-minute throttle alerts.
Pillar C
The Microsoft Fabric Capacity Metrics App: A Complete Guide
Annotated walkthrough of the Fabric Capacity Metrics app: Compute (14-day), Storage (30-day), Throttling, Timepoint, and Health pages — limits and gaps.
Pillar A
Fabric Capacity Sizing Guide: Pick the Right F-SKU First Time
Fabric capacity sizing is a smoothing question, not peak CPU. Heuristic table maps BI-only, ETL-heavy, data-science, and mixed loads to a starting F-SKU.
Pillar C
The Fabric Chargeback App vs True Cost Attribution: Limits & How to Go Deeper
The Fabric Chargeback app refreshes daily, masks service-principal workloads, and stops at item grain. Here's what it covers and how to go deeper.
Pillar B
Controlling Copilot & AI Costs in Fabric: The New Billing Meters
Copilot draws from your Fabric CU pool with no native kill switch. Learn what the Copilot-and-AI meter costs and how to stop a runaway-AI loop.
Pillar C
Fabric Cost Monitoring: Native Tools vs Purpose-Built (Honest Comparison)
Five Fabric cost monitoring tools compared — Capacity Metrics app, Chargeback, FUAM, Fabric Cost Analysis, and SpendWeave Pro. Honest capability matrix.
Pillar C
Fabric Capacity Alerts That Actually Fire: Real-Time Hub Over Metrics App
Fabric's Capacity Metrics app has no built-in alerts. One Activator rule on Real-Time hub capacity events fires the moment throttling starts.
Pillar C
Fabric Capacity Metrics Retention: 14-Day Compute Limit, 30-Day Storage — and How to Keep History
Fabric Capacity Metrics keeps 14 days of compute detail and 30 days of storage. No setting extends either. Retention matrix and extraction pattern inside.
Pillar B
The Fabric Pause-and-Resume Trap: How Pausing Bills You
Pausing a Fabric capacity carrying smoothed debt settles the overage immediately at PAYG rates. On reserved, you pay twice. Mechanism and math.
Pillar C
PAYG Overage Billing (2026): What It Changes for Your Forecast
Fabric PAYG overage charges at 3× the PAYG rate on a separate Azure meter. Reserved capacity does not cover it — here is what that means for your forecast.
Pillar B
Fabric Pipeline Minute-Rounding: How Sub-Minute Copy Activities Can Cost up to 12x Their Runtime
Fabric pipeline copy activities bill per minute, rounded up. A 14-second run bills as 60 seconds — ~328% inflation. Full penalty curve and fixes.
Pillar A
Reserved vs Pay-As-You-Go: Which Fabric Capacity Model Saves More
A Fabric reservation costs 59.49% of PAYG — but only wins if uptime exceeds ~59.5% of the month. Breakeven math, worked F32 example, and the pause-trap.
Pillar A
One Big Capacity vs Many Small: Fabric Topology Trade-offs
Pooling workloads into one Fabric capacity cuts cost per CU but creates a capacity-wide blast radius. Here is how to pick the right topology for your org.
Pillar B
Smoothing, Bursting & Carry-Forward Debt: The Physics of Fabric Compute
How Fabric's 30-second timepoints, 10-minute interactive window, and 24-hour background smoothing create CU debt and trigger the staged throttle cascade.
Pillar C
Fabric Throttling Explained: Why One Query Slows Your Whole Tenant
Fabric throttling is staged across future-capacity time windows, not utilization percentages. One workload's debt blocks every user on the shared capacity.
Pillar C
Real-Time Throttling Triage When the Metrics App Can't Keep Up
The Capacity Metrics app lags 10–15 min — live triage runbook for an active throttle: what to read, in what order, and why pausing is a costly trap.
Pillar B
When Fabric Pipelines Cost up to 11x Self-Hosted-IR ADF (and When They Don't)
Fabric pipelines bill on duration vs ADF self-hosted-IR. A side-by-side table shows when the cost gap reaches ~10x and when it narrows sharply.
Pillar C
Workload Isolation & the Blast Radius: When to Split Capacities
No native per-workspace CU isolation exists in Fabric. One runaway job throttles your whole capacity. Size the blast radius and decide when to split.
Pillar B
The Hidden Costs of Microsoft Fabric (with the Math)
Seven hidden Microsoft Fabric cost categories, each with exact dollar math from published rates — idle compute, pause trap, pipelines, OneLake, and more.
Pillar C
Microsoft Fabric Capacity Monitoring: The Complete Guide
Monitor Fabric capacity with the free Capacity Metrics app — 14-day compute, 30-day storage, and the gaps. The native-vs-gap matrix and what fills it.
Pillar A
Microsoft Fabric F-SKU Pricing Explained (F2-F2048)
Every F-SKU from F2 to F2048 — PAYG and reserved prices, what each tier unlocks (Copilot, free viewers, mirroring), and which rungs to skip.
Pillar A
The F64 Licensing Cliff: When Fabric Stops Needing Power BI Pro
At F64 ($8,409.60/mo PAYG), Power BI viewers need no Pro license. Below F64, every viewer costs $14/mo. Per-SKU breakeven table inside.
Pillar A
Microsoft Fabric Pricing & Capacity Planning Guide
Microsoft Fabric pricing starts at $262.80/mo (F2) and hits the F64 licensing cliff at $8,409.60/mo. The full F-SKU table plus a tier-picking decision flow.
Pillar A
OneLake Storage Costs Explained: The Hidden $0.023/GB
OneLake costs $0.023/GB-month, separate from your F-SKU. Soft-delete, BCDR geo-replication, and mirroring-pause overages turn 1 TB logical into 3+ TB billed.
Pillar C
Predicting Fabric Throttling Before It Happens
Track carry-forward debt slope to forecast Fabric throttle onset before users hit errors — a debt-trajectory method with a worked F32 example.
Pillar B
How to Reduce Microsoft Fabric Costs: The Hidden-Waste Playbook
The fastest way to cut a Fabric bill: kill idle PAYG capacity (an idle F64 burns ~$8,410/mo). Plus a prioritized waste checklist — and the pause trap to avoid.
Pillar B
Right-Sizing Your Fabric Capacity: Find & Recover Wasted CUs
Right-size a Microsoft Fabric capacity by reading smoothed load, not peak CPU. Worked example shows how to recover real dollars from a misread ribbon.
Pillar A
What Is a Capacity Unit (CU)? How Microsoft Fabric Billing Works
A Fabric Capacity Unit (CU) is the single billing token behind every Fabric workload. At $0.18/CU-hour, F64 costs $8,409.60/mo — here is the exact math.