metering.vcpu_estimate_per_query ​

The metering.vcpu_estimate_per_query view contains information about the queries that use the most vCPU on the system. On every metering execution, the top 100 queries based on maximum execution time (run_ms) will get loaded and stored in this view.

The vcpu_estimate_seconds is calculated by using the formula:

vcpu_estimate_seconds = (run_ms – wait_run_cpu_ms) x No. of workers x cores per worker

• run_ms: Time required for query to get executed. It tells us how long the query was running on the longest-running worker node. This value represents the total time the query spent in the run state.

• wait_run_cpu_ms: Duration needed for processing the query.

Note: The value (run_ms – wait_run_cpu_ms) is rounded up.

The run_ms value is NULL before the query enters the run state, then it is initialized to 0.

This view can be used to understand if the cluster capacity is properly utilised.