Understanding Performance dashboard vs. capacity metric dashboard

[Anonymous]

Hello everyone,

since the 1st of Feb 2025 the SQL database compute and storage is being charged to the Fabric capacity. This let to a lot of throttling in our capacity. We were at a F16 capacity, but upgraded to F32 to prevent the throttling. 

But I was wondering how the SQL database was operating so I checked the Performance Dashboard (Fabric SQL database) and there I see that we are not even using 0.3 CPU, with a status of normal.   

While the capacity metrics show that we are exceeding our SKU. How is it possible? 

regards,
Mark

[Anonymous]

Hi @nilendraFabric ,

the screenshots are from the F32 SKU. However the question remains the same. 

There are other processes running on the capacity. However between 3 am and 4 am, SQL is the only activity classified as interactive. The capacity is processing around <20% on background activities at that time.  

[nilendraFabric]

Hello @Anonymous 

could you please confirm if Sql database is the only workload utilising the F16?

SQL Database uses only a portion of Fabric’s total capacity (1 Fabric CU = 0.383 SQL vCores). A Fabric F64 SKU (64 CUs) provides ~24.5 SQL vCores, leaving ~39.5 CUs for other workloads like Power BI, Data Engineering, or KQL databases

Throttling occurs at the capacity level, not per workload. Even with low SQL CPU usage, concurrent workloads (e.g., Spark jobs, Power BI reports) can collectively exceed capacity limits

Most SQL operations are classified as interactive in Fabric and smoothed over 5-minute windows. Short, frequent queries or system tasks (e.g., metadata updates) can spike CU usage without significantly impacting SQL CPU

this is mainly for guidance. 

thanks