Your GKE cluster hosts a web application that experiences sporadic traffic spikes. The application must scale to maintain uptime during these spikes, but you also wish to conserve resources when demand is low. What is the best autoscaling configuration for this scenario?