You have deployed a data processing application in a GKE cluster. The application processes large datasets during business hours, causing spikes in CPU and memory usage. Outside business hours, the workload is minimal, and you want to optimize costs by scaling down unused resources. How can you achieve this?