Professional Cloud ArchitectManage and provision solution infrastructureMedium
A research institution is running computationally intensive simulations using numerous virtual machines on Compute Engine. They need to minimize the cost of these long-running, fault-tolerant workloads that can be interrupted and restarted without significant impact. Which type of Compute Engine instance is most cost-effective for this use case?
- AStandard Compute Engine instances
- BPreemptible VMs
- CCustom machine types
- DSole-tenant nodes
Show answer & explanationAnswer & explanation
Correct answer: B. Preemptible VMs
Preemptible VMs are significantly cheaper than standard instances but can be preempted by Compute Engine if resources are needed elsewhere. They are ideal for fault-tolerant, batch, or computationally intensive workloads that can handle interruptions and restarts, making them highly cost-effective for the described research simulations.
Why the other options are wrong
- A. Standard instances are more expensive as they offer guaranteed uptime, which is not strictly required for fault-tolerant, interruptible workloads.
- C. Custom machine types allow tailoring CPU/memory but don't inherently offer the cost savings or preemption characteristics needed for this specific scenario.
- D. Sole-tenant nodes provide dedicated physical servers, offering isolation and specific licensing benefits, but are significantly more expensive and not for cost minimization of interruptible workloads.
Preemptible VMs
Compute Engine instances that are significantly cheaper than standard VMs but can be terminated by Google Cloud at any time (preempted) when resources are needed elsewhere.
- Up to 80% cheaper than standard VMs
- Maximum run time of 24 hours (can be shorter due to preemption)
- Ideal for fault-tolerant, batch, or stateless workloads
Memory trick: Preemptible VMs save money if you can handle a little pause.