Forum Discussion

kashif_m's avatar
kashif_m
Regular Visitor
2 months ago

Community tool: Fabric Capacity Simulator — model workloads, throttling & cost

Hi everyone,

I built a free community tool to help with Fabric capacity planning, especially when you're deciding on F-SKU size, autoscale, or capacity overage, and want to understand throttling impact before changes go live.

The problem it solves

If you manage Fabric capacity, you've probably faced questions like:

  • Will F64 vs F128 handle our overnight ETL and morning report rush?
  • What happens when interactive and background workloads overlap — delay, reject, or both?
  • Is autoscale worth it for our usage pattern?
  • When does capacity overage (preview) make sense vs right-sizing the SKU?
  • What might daily cost look like under PAYG, reserved, autoscale, and overage?

The Capacity Metrics app shows what happened in your tenant. This simulator helps with what-if model workloads over 24 hours and see utilisation, throttling stages, rejection, and estimated cost live as you adjust settings.

It is aligned with Microsoft capacity concepts (bursting, smoothing, progressive throttling) documented here:
https://learn.microsoft.com/en-us/fabric/enterprise/throttling

Try it (live)

https://fabric-capacity-simulator.onrender.com/

No sign-in required. Open a preset scenario, tweak workloads, and watch the chart update in real time.

What it includes

  • Live 24-hour simulation — 30-second timepoints with bursting, smoothing, and throttling behaviour
  • Realistic preset scenarios — enterprise-style mixes (overnight Spark ETL, semantic refresh, morning report rush, ad-hoc notebooks, etc.)
  • Per-workload schedule — gantt lanes on the chart with colour-coded demand
  • Capacity settings — Autoscale max, capacity overage limit
  • Cost estimates — PAYG, reserved, autoscale, and overage 
  • Learn tab — visual guide to capacity management concepts
  • Export — download scenario data for your own analysis

Quick start

  1. Open the simulator and pick a scenario (e.g. Morning Peak — Enterprise Mix).
  2. Review Overview KPIs: peak util, throttled time, rejected CU-s, estimated daily cost.
  3. Toggle autoscale or overage and compare the timeline.
  4. Add or edit workloads with the + button to model your own pattern.
  5. Visit the Learn tab for background on how capacity behaviour works.




Disclaimer

This is a community-built tool. It is not affiliated with, endorsed by, or supported by Microsoft. Simulation and costing are estimates for planning and education, always validate against your tenant's Capacity Metrics app and Azure billing before making production decisions.

Feedback welcome

I'd love your feedback:

  • Do the scenarios feel realistic for your environment?
  • What workloads or features would you add?
  • Any behaviour that doesn't match what you see in Metrics app?

Need help applying this to your tenant?

If you want a capacity sizing review or help translating real Metrics data into a sizing recommendation, I'm happy to chat:

Hope this is useful for admins and architects sizing Fabric capacity. Enjoy exploring, and please share what you think.


Kashif

admin
capacities
governance
workspaces

2 Replies

  • awehro's avatar
    awehro
    Frequent Visitor
    1 month ago

    Hi Kashif, 
    Nice work. 
    This is one of the more thorough Fabric capacity tools shared in this forum. A couple of questions and one suggestion:

    1. Integration with Capacity Metrics App: Have you considered a path to import actual utilization data from Capacity Metrics App into the simulator, so scenarios can be validated against real tenant history instead of starting from presets alone?

    2. Your experience with Capacity Metrics App: Do you use it as-is for your own capacity work, or did you build a custom solution to cover gaps in what it reports? We've repeatedly activated it across our capacities and found it doesn't consistently surface complete data for all capacities in the tenant. Curious whether you've run into the same issue and how you've worked around it.

    3. Suggestion for the simulator: It could be useful to let users seed workload assumptions from tenant metrics such as Power BI user counts, dataset/report counts, or refresh frequency, rather than only choosing from fixed presets. That would put the initial scenario setup closer to a given tenant's actual footprint before manual tuning.

    Thanks for sharing this; I will be testing it against our own workload mix.

  • v-achippa's avatar
    v-achippa
    Community Support
    1 month ago

    Hi kashif_m,

     

    Thank you for reaching out to Microsoft Fabric Community.

     

    Thank you for sharing the tool. This will be really helpful for the community. We will keep this thread open for the community members to explore and ask questions and share their feedback on this.

     

    Thanks and regards,

    Anjan Kumar Chippa