Fabric Data Warehouse Cost

[Flubis]

Hello, 

I have been searching for the true cost of using a mirrored Snowflake (Fabric Data Warehouse) within the P1 SKU. I have come across many articles stating that it is essentially "free" on the Fabric side if your data is within 25 GB in size (e.g., Radacad and various other community-generated posts). While I would like to believe this is true, I suspect there are costs on the Fabric side that I should be aware of.

I understand that most of the costs will be incurred on the Snowflake side, particularly for egress and Data Warehouse activation. However, I would like to know what the associated costs are for running a mirrored Snowflake Data Warehouse in Fabric.

[Flubis]

Thanks all!!

[Poojara_D12]

Hi @Flubis 

Using a mirrored Snowflake Data Warehouse in Microsoft Fabric within the P1 SKU has these cost considerations:

1. Fabric Costs:

   • Under 25 GB: Free and included in the P1 SKU.
   • Over 25 GB: May consume Fabric capacity (CCUs), impacting other workloads.

2. Snowflake Costs:

   • Egress Costs: Charged for data transfer out of Snowflake.
   • Compute Costs: For queries or transformations triggered in Snowflake.

3. Recommendation:

   • Keep mirrored data under 25 GB to minimize costs.
   • Monitor Fabric capacity usage to optimize workloads.

Snowflake costs are separate and depend on your usage.

Did I answer your question? Mark my post as a solution, this will help others!
If my response(s) assisted you in any way, don't forget to drop me a "Kudos" 🙂

Kind Regards,
Poojara
Data Analyst | MSBI Developer | Power BI Consultant
Consider Subscribing my YouTube for Beginners/Advance Concepts: https://youtube.com/@biconcepts?si=04iw9SYI2HN80HKS 

[V-yubandi-msft · Accepted solution]

Hi @Flubis ,

Thank you for connecting with the Microsoft Fabric Community.

Fabric does not impose network data ingress fees for mirroring into One Lake. This means that when your Snowflake data is replicated into One Lake, there are no associated mirroring costs.

However, there are compute and cloud query costs on the Snowflake side. These include virtual warehouse compute costs and cloud services compute costs incurred during the data mirroring process.

• When mirroring data from Snowflake to Fabric, compute costs are incurred on the Snowflake side due to data extraction. These costs are influenced by the size of the virtual warehouse and the duration of query execution. Larger warehouses and longer queries result in higher costs. Operations by Snowflake's cloud services, such as managing and optimizing queries, also add to the overall compute costs.
• Additionally, if Fabric frequently queries Snowflake to detect data changes for replication, each of these queries will consume compute resources on the Snowflake side.

To manage and potentially minimize these costs, consider the following:

• Fine-tune data extraction queries to boost performance. This cuts down the runtime of the virtual warehouse, saving costs.
• Plan mirroring operations during off-peak hours or do it less often, depending on how fresh you need your data. This keeps compute costs in check.
• Keep an eye on the virtual warehouse's usage and performance. Snowflake offers tools to analyse performance and spot ways to save costs.

For more information and clarity, please refer to the official Microsoft documentation on Snowflake mirror costs and configuring Snowflake activity mentioned in the links below.

 Microsoft Fabric Mirrored Databases From Snowflake - Microsoft Fabric | Microsoft Learn.

 Configure Snowflake in a copy activity - Microsoft Fabric | Microsoft Learn.

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

Additionally, I would be grateful for a 'Kudos' if you found my response helpful.

[GilbertQ]

Hi @Flubis 

When people talk about the free portion when using fabric that relates to the data that is being stored, you will not incur any cost. After that, if you are querying data from the Snowflake Mirror, you will incur cost or capacity units, which will be used when queries being run. 

What I would recommend doing is to create the snowflake mirror and then run a few queries. Once that is done, go do the metrics app and you can then see how many capacity units is being consumed.