The Compute Engine feature you’re asking about is Autoscaling.

Quick Scoop: Direct Answer

In Google Cloud Compute Engine, the feature that allows virtual machines (VMs) to be automatically added or subtracted from an application based on load metrics (like CPU, load balancer utilization, or custom monitoring metrics) is called Autoscaling.

What Autoscaling Actually Does

Autoscaling works with managed instance groups and watches metrics such as:

  • CPU utilization percentage.
  • HTTP(S) load balancing utilization.
  • Cloud Monitoring metrics (for example, latency, queue length, Pub/Sub backlog).

When the metric goes above a target threshold, autoscaling creates more VMs to handle the load, and when it falls below the target, it gradually deletes VMs to save cost.

Mini Example Story

Imagine you run a web app on a managed instance group:

  • During normal hours, you might run 3 VMs.
  • At peak time, traffic spikes and CPU usage jumps over your configured target (say 60–70%). Autoscaling detects this and automatically adds more VMs (for example up to 10) so users don’t feel slowdowns.
  • Late at night, traffic drops, metrics fall below target, and autoscaling removes extra VMs so you are not paying for idle capacity.

This is exactly the “VMs added to or subtracted from an application based on load metrics” behavior your question describes.

Key Facts at a Glance

  • Feature name: Autoscaling.
  • Scope: Works with managed instance groups in Compute Engine.
  • Triggers: CPU, load balancer utilization, or Cloud Monitoring metrics.
  • Goal: Maintain performance while optimizing cost by automatically scaling out and in.

TL;DR: The Compute Engine feature that automatically adds or removes VMs based on load metrics is Autoscaling.

Information gathered from public forums or data available on the internet and portrayed here.