A nightly batch job uses many VMs, is fault-tolerant to instance termination, and VM costs are high. What should you do to reduce cost?

Preemptible VMs are significantly cheaper than standard VMs and are ideal for fault-tolerant, batch processing workloads. They can be terminated by Google Cloud at any time, but since the job is fault-tolerant, it can handle these interruptions. Running a simulated maintenance test helps ensure the application gracefully handles preemptions before fully committing. Using regular N1 Standard VMs (without preemption) would not provide the same cost savings. Managed instance groups are good for scaling and self-healing, but don't inherently reduce costs for individual instances compared to preemptible options. Switching from N2 to N1 VMs might offer some savings, but preemptible VMs offer a much greater cost reduction for this type of workload.