You are designing a batch data-processing pipeline whose jobs are fault-tolerant and can be interrupted and restarted without impact. To minimize compute costs for this workload, which Compute Engine option should you plan to use?