A customer has an application running in virtual machines. They are migrating this application to Google Cloud. They have previously had scaling issues when on-premises as VMs had to be pre-allocated. Capacity planning was repeatedly off mark - it's either too many VMs or too less. They want to match the capacity to demand while keeping the application running always. They don't have the time or budget to re-architect the systems using containers and Kubernetes at the mo-ment. What would be your recommendation?