toolbox
Home / Startups & Digital Business / Serverless Function Utilization Rate Estimator

Serverless Function Utilization Rate Estimator

The Serverless Function Utilization Rate Estimator compares observed average and peak concurrent executions with a selected concurrency capacity. It provides separate average and peak utilization percentages, making it easier to distinguish a consistently underused reservation from a workload that looks quiet on average but approaches its limit during bursts.

Teams can use this metric when reviewing provisioned concurrency, reserved limits, or an internal capacity target. Utilization by itself does not measure response time, cold starts, or throttling, so it should be interpreted together with service-level and scaling data.

Concurrency utilization

concurrency
concurrency
concurrency
Result
—
Average concurrency utilization
Peak utilization—
Average spare capacity—
Peak spare capacity—

1. Choose the capacity denominator
Enter the concurrency capacity you want to evaluate, such as provisioned concurrency or an internal operating target.

2. Enter average concurrency
Use an average measured over a representative period that matches the capacity configuration.

3. Enter peak concurrency
Use the highest observed or planned concurrency for the same period and workload scope.

4. Compare average and peak utilization
Average utilization shows how intensively capacity is used overall, while peak utilization reveals how close bursts come to the selected capacity.

5. Inspect spare capacity
Positive spare capacity indicates unused headroom; a negative peak value means observed demand exceeded the selected capacity figure.

Average utilization (%) = Average concurrency ÷ Capacity × 100

Peak utilization (%) = Peak concurrency ÷ Capacity × 100

Spare capacity is capacity minus observed concurrency. Utilization can exceed 100% if the entered demand is greater than the selected capacity; this may be meaningful when comparing observed demand with a smaller reservation or target rather than a hard execution limit.

What the result means

The main result shows the share of selected concurrency capacity used on average during the measurement period.

A low average with a high peak may indicate bursty demand, so sizing decisions should not rely on the average alone.

Given
Average concurrency = 42
Peak concurrency = 115
Selected capacity = 150

Calculation
Average utilization = 42 ÷ 150 × 100 = 28.0%
Peak utilization = 115 ÷ 150 × 100 = 76.7%
Average spare capacity = 108
Peak spare capacity = 35

Result
Average utilization = 28.0%.

The workload uses less than one-third of the selected capacity on average but reaches more than three-quarters of it at peak.

Can utilization be above 100%?

Yes, if the observed concurrency entered is greater than the selected comparison capacity. Whether that represents actual throttling depends on what the capacity figure represents in your platform.

What measurement window should I use?

Use a period long enough to represent normal behavior but short enough that the capacity configuration and workload have not materially changed. Keep the window consistent when comparing trends.

Why track both average and peak utilization?

Average utilization captures general efficiency, while peak utilization captures burst pressure. A workload can have low average use and still require substantial headroom for short peaks.

Does low utilization always mean I should reduce capacity?

No. Spare capacity may be intentional for latency, burst tolerance, or service-level objectives. Review performance and scaling behavior before changing a reservation or limit.

How is utilization different from capacity needs?

Utilization evaluates how current or selected capacity is being used. Capacity-needs estimation starts from traffic rate and execution duration to calculate how much concurrency may be required.