Forum Discussion

andremaarinho's avatar
andremaarinho
New Member
10 months ago
Solved

Shared Semantic Model Costs

Hey all!! 

 

In my company's current setup, we have multiple Power BI reports that consume data from the same semantic model. We’re now centralizing everything into a single semantic model file. However, we have some concerns regarding potential additional costs.

While reading the Fabric documentation, I came across the metric "live connections (per second)". We’re currently using an F8 capacity, and I’d like to understand exactly what this metric represents.

 

 

 

Does “live connections (per second)” refer to each time a user accesses a dashboard, or is it counted every time the user interacts with it (e.g., using slicers, filters, or visuals)?

 

Our dashboards are embedded, and each user interaction triggers a query to the model. The reports connect to the semantic model via DirectQuery, but the semantic model itself is in Import mode. That’s why I’m trying to understand whether every interaction would count as a new live connection and potentially impact capacity costs.

 

It'd be very much appreciated if you all could help me understand this 🙂

 

Thank youu

  • Hello andremaarinho,

     

    In Fabric capacity metrics, “Live connections (per second)” measures how many concurrent query sessions (connections) are established from reports or dashboards to semantic models per second.

    It is not a billing or cost metric — it’s a performance indicator showing query concurrency and load patterns on your capacity.


    🔹 What counts as a “live connection”?

    • A live connection occurs whenever a report establishes a session with a semantic model hosted in your Fabric capacity.

    • This usually happens once per report render (initial load or refresh) — not per interaction.

    • Inside that connection, multiple queries (DAX or MDX) are executed per visual, slicer, or filter, but these are all part of the same active session.

    So:

    No, you are not charged per live connection.

    Fabric capacity cost is fixed based on the SKU (F2, F4, F8, F64, etc.).
    What this metric reflects is resource consumption (CPU, memory, and concurrency), which determines whether your capacity is adequate for your workload.

    What can impact cost:

    • If you consistently hit high concurrency, query queueing, or refresh delays, you might need to scale up (F16) or autoscale to maintain performance.

    • But every slicer click does not create new billing activity.

     

    Documentation : 

    https://learn.microsoft.com/en-us/fabric/enterprise/licenses#capacity-units

    https://learn.microsoft.com/en-us/fabric/enterprise/metrics-app

    https://learn.microsoft.com/en-us/fabric/enterprise/metrics-app-compute-page

     

    Hope it can help you !

    Best regards,

    Antoine

3 Replies

  • Hello andremaarinho,

     

    In Fabric capacity metrics, “Live connections (per second)” measures how many concurrent query sessions (connections) are established from reports or dashboards to semantic models per second.

    It is not a billing or cost metric — it’s a performance indicator showing query concurrency and load patterns on your capacity.


    🔹 What counts as a “live connection”?

    • A live connection occurs whenever a report establishes a session with a semantic model hosted in your Fabric capacity.

    • This usually happens once per report render (initial load or refresh) — not per interaction.

    • Inside that connection, multiple queries (DAX or MDX) are executed per visual, slicer, or filter, but these are all part of the same active session.

    So:

    No, you are not charged per live connection.

    Fabric capacity cost is fixed based on the SKU (F2, F4, F8, F64, etc.).
    What this metric reflects is resource consumption (CPU, memory, and concurrency), which determines whether your capacity is adequate for your workload.

    What can impact cost:

    • If you consistently hit high concurrency, query queueing, or refresh delays, you might need to scale up (F16) or autoscale to maintain performance.

    • But every slicer click does not create new billing activity.

     

    Documentation : 

    https://learn.microsoft.com/en-us/fabric/enterprise/licenses#capacity-units

    https://learn.microsoft.com/en-us/fabric/enterprise/metrics-app

    https://learn.microsoft.com/en-us/fabric/enterprise/metrics-app-compute-page

     

    Hope it can help you !

    Best regards,

    Antoine