What is the autoscaling target value?

The target value (or target utilization) is the metric level at which the autoscaler is satisfied — it will add replicas when the metric exceeds the target and remove them when it falls below. For Kubernetes HPA, this is often expressed as a percentage (e.g., 60 for 60% CPU). For AWS Target Tracking, it is the raw metric value (e.g., 1000 requests per target).

How does Kubernetes HPA calculate desired replicas?

Kubernetes uses: desiredReplicas = ceil(currentReplicas × (currentMetricValue / desiredMetricValue)). For example, 3 replicas at 80% CPU with a target of 60% → ceil(3 × 80/60) = ceil(4) = 4 replicas.

Why set the target below 100%?

Setting the target below 100% (commonly 60–80%) leaves headroom for traffic spikes. If the target equals 100%, a sudden burst can saturate all replicas before new ones are ready, causing latency or errors.

What is the difference between HPA and AWS Target Tracking?

Both use the same ceil(current × observed/target) formula. Kubernetes HPA typically uses percentages for CPU/memory or raw values for custom metrics. AWS Target Tracking uses raw CloudWatch metric values (e.g., ALBRequestCountPerTarget).

What does "back-calculate target from desired replicas" mean?

If you already know how many replicas you want at the current load, you can compute target = currentMetricValue × (currentReplicas / desiredReplicas). This tells you what target to set in the autoscaler config.

Autoscaling Target Calculator

Name: Autoscaling Target Calculator
Availability: InStock
Author: Nham Vu

Enter your current replicas, observed metric, and desired utilization to compute the recommended target value and predicted replica count.

Autoscaling Parameters

Calculation Mode

Given a target utilization, predict how many replicas will run.

Autoscaler Type

Current Replicas

Observed Average Metric Value (% per pod)

Desired Target Value (% — e.g. 60)

Min Replicas

Max Replicas

Result

Fill in the parameters and click Calculate.

Summary

Enter your current replicas, observed metric, and desired utilization to compute the recommended target value and predicted replica count.

How it works

Select the autoscaling type: Kubernetes HPA (resource utilization) or AWS Target Tracking (request count, CPU %, etc.).
Enter the current number of running replicas or instances.
Enter the current observed metric value (e.g., average CPU % or requests per second per pod).
Enter your desired steady-state utilization (e.g., 60% CPU).
The calculator outputs the recommended target value to set in your autoscaling policy and the predicted replica count.
Use the replica boundary fields to clamp results within your min/max replica limits.

Use cases

Tune Kubernetes HPA target CPU/memory utilization to avoid over- or under-provisioning.
Configure AWS Target Tracking policies for EC2 Auto Scaling groups.
Predict how many pods will spin up during a traffic spike given a target metric.
Back-calculate the target value needed to achieve a desired number of replicas at current load.
Validate autoscaling config before a load test to confirm the scaler behaves as expected.
Determine safe headroom: set target below 100% so replicas have burst capacity.

Frequently Asked Questions

Last updated: 2026-06-11 · Reviewed by Nham Vu