Your organization wants to optimize costs for a set of non-critical, short-lived Compute Engine instances. These instances should automatically terminate after their work is done and should be very cost-effective. What would you recommend?

*Explanation:* _Use preemptible instances. A. Incorrect: f1-micro types are not necessarily the most cost-effective for short-lived workloads. B. Incorrect: Custom machine types could be over-provisioned for the task. C. Incorrect: On-demand instances are more expensive than necessary. D) Correct: Preemptible instances are cost-effective and designed for short-lived tasks. E) Incorrect: Stackdriver alerting adds unnecessary complexity. F) Incorrect: Sole-tenant nodes are not designed for cost optimization._