Statement CFU Metrics in Confluent Cloud for Apache Flink
A statement CFU metric reports the Confluent Flink Units (CFUs) that an individual Flink SQL statement consumes over time, so you can attribute compute pool usage to specific statements in Confluent Cloud for Apache Flink®. Use these metrics to monitor per-statement cost, track how a compute pool’s resources are distributed, and identify high-consumption statements to optimize.
How to use statement CFU metrics
The statement CFU metrics give you insights into the resources that individual statements running inside your compute pools consume.
Specifically, the statement CFU metrics enable you to:
Monitor individual statement usage: Accurately measure the number of CFUs each statement consumes over time. This metric is available for all types of statements that you submit in Confluent Cloud for Apache Flink.
Track resource distribution: Understand how Confluent Cloud for Apache Flink distributes the resources in a compute pool among the statements that run in the compute pool.
Identify high-consumption statements: Pinpoint which statements consume the most CFUs, enabling you to optimize the statement’s Flink SQL code or adjust the resources available to this statement.
By monitoring statement-level CFU consumption, you can make informed decisions about your Flink application’s cost efficiency and resource utilization.
You can’t set minimum or maximum CFU limits on individual statements, but maximum CFU limits are configurable at the compute-pool level.
Where to view statement CFU metrics
The statement CFU consumption metrics are available to view in the statements summary table and in the statement side panel.
Statements summary table: Get an overview of CFU consumption for all your statements directly within the statements summary table. This provides a quick way to identify the most resource-intensive statements.
Statement side panel: For a deeper dive into a statement’s resource usage, open the statement side panel, where you can see the current CFU consumption and a time-series chart that shows how the statement’s CFU consumption has evolved over time.
How UDF resource consumption is represented
The statement CFU metric shows the resources that your SQL statements consume, along with the resources that any UDF instances the statement might invoke consume.
Resources that individual UDF instances consume sometimes appear as fractional CFU values, because Confluent Cloud for Apache Flink can consolidate (or roll) multiple UDF instances into a single CFU of resources. Confluent Cloud for Apache Flink can combine up to four instances of a UDF into one CFU.
The distribution of CFUs among UDFs in a compute pool is flexible. Confluent Cloud for Apache Flink can roll four UDF instances across different statements into a single CFU, as long as the statements are in the same compute pool. Also, Confluent Cloud for Apache Flink can consolidate different UDF functions and their instances into a single CFU, as long as they are in the same compute pool.
Understanding differences between CFUs for compute pool and statement
When monitoring resource consumption in Confluent Cloud for Apache Flink, you might observe minor differences between your compute pool CFU metrics and the aggregated sum of your statement CFU metrics. These discrepancies are expected and are caused by rounding.
If your statements use UDFs, you might see a maximum discrepancy of 2 CFUs between the compute pool CFU metrics and the total sum of your statement CFU metrics.
For statements not using UDFs, the maximum expected discrepancy between the compute pool CFU metrics and the total sum of your statement CFU metrics is 1 CFU.
Note
Confluent bills you based on the compute pool CFU metrics, not on the summed total of individual statement CFU metrics.