You must run a single binary on VMs and autoscale based on CPU; policy requires using VMs directly. For efficient, fast scaling, what should you do?

For efficient, fast scaling of VMs based on CPU, creating an instance template and using it in a managed instance group (MIG) with autoscaling is the best solution. An instance template defines the VM configuration, and a MIG uses this template to create identical VMs. Autoscaling within the MIG automatically adds or removes VM instances based on CPU utilization, meeting the requirement for fast, efficient scaling. Using Google Kubernetes Engine (GKE) is incorrect because the policy explicitly requires using VMs directly, not containers orchestrated by Kubernetes. Scaling on a time schedule is incorrect because the requirement is to autoscale based on CPU, not a fixed schedule. Relying on third-party tools is less efficient and integrated than Google Cloud's native autoscaling capabilities, which are designed for this exact scenario.