CU consumption of Fabric Database

Question

Dear all,&nbsp;&nbsp;We are migrating some of our components to SQL DB.&nbsp;&nbsp;Architecture : F2&nbsp;The use case :&nbsp;- In the course of our pipeline, we fetch and update some data / flags / timestamp from config tables that we used to host in a WH (to benefit from SQL Script). As Delta is not the best option when it comes to concurrent updates, we want to move to a Fabric DB.&nbsp;- In term of features it works perfectly.&nbsp;- But we are a bit concerned by the data displayed in the Fabric Capacity Metric regarding the usage.&nbsp; (Interactive Non Billable ?? )&nbsp;&nbsp;It seems very high to us and we are reluctant to go with this option on a bigger scale, if we have to fear hug impact in the performances.Do you have any explanation on the consumption ? Any reference ? I did not find anything in the documentation.&nbsp;&nbsp;&nbsp;

nilendrafabric · Accepted Answer

Hello&nbsp;RomainDomart1&nbsp;&nbsp;Fabric classifies pipeline-triggered activities as interactive operations when they:• Use T-SQL DML statements (INSERT/UPDATE/DELETE) on config tables• Query DMVs/system catalog views&nbsp;• Contain cross-database operations&nbsp;Your F2 SKU (2 CUs ≈ 1 vCore) has limited interactive operation capacity:&nbsp;Interactive | ~3,600 CU-seconds/hrBackground | ~1,800 CUsec/hr&nbsp;https://github.com/MicrosoftDocs/fabric-docs/blob/main/docs/data-warehouse/how-to-observe-utilization.md&nbsp;The “Interactive Non-Billable” label appears because:• Fabric SQL Database remains free until 2025-01-01&nbsp;• System-generated operations (pipeline coordination, statistics updates) don’t incur charges&nbsp;hope this helps&nbsp;now to improve performance and optimise cost try these&nbsp;Replace frequent small updates with batch operations -- merge&nbsp;Use the DMV query to track expensive operations&nbsp;Thanks

Forum Discussion

CU consumption of Fabric Database

2 Replies

Related content

Fabric Capacity Consumption

Regarding Data agent in Fabric consumption

Fabric Mirrored Databases

Memory consumption metrics in Fabric SQL Database

Less resource consumption