You are managing a memory-intensive, stateless application that needs to handle variable workloads in a GKE cluster. The application needs to scale quickly in response to workload changes to maintain performance. Which autoscaling strategy should you implement to meet these requirements?