A GKE data-processing app requires nodes with GPUs, local SSDs, and 8 Gbps network; containers must launch with the latest images. How should you efficiently provision infrastructure and manage deployments?

The correct answer leverages Cloud Build and Terraform for efficient, automated infrastructure provisioning and deployment. Terraform allows you to define the GKE cluster with specific node pools (including GPUs, local SSDs, and network bandwidth) as infrastructure as code, ensuring consistency and repeatability. Cloud Build can then execute this Terraform configuration, and also build and push the latest container images, ensuring deployments always use the most current versions. Incorrect options: Compute Engine startup scripts are not ideal for managing complex GKE cluster infrastructure and container deployments. gcloud commands are manual and less scalable for infrastructure as code. While GKE autoscaling is useful for managing container resources, it doesn't address the initial provisioning of the specialized GKE cluster with GPUs, SSDs, and specific networking, nor does it handle launching with the latest images. Dataflow is a managed service for executing data processing pipelines, not for provisioning GKE infrastructure or managing container deployments. Cloud Scheduler is for scheduling tasks, but not for infrastructure provisioning itself.