A company needs to plan compute resources for a nightly data indexing task on Google Cloud. The task runs once per day for approximately 4 hours, periodically saves checkpoint state to Cloud Storage, and can safely resume without data loss if pre-empted. The primary goal is to minimize compute infrastructure costs. Which compute deployment strategy should the cloud engineer select?
- Provision Compute Engine instances utilizing Spot VMs for the batch task execution.Answer
- BDeploy the indexing logic as a Cloud Function triggered nightly by Cloud Scheduler.
- CProvision standard Compute Engine VMs attached to a 3-year Committed Use Discount contract.
- DDeploy the workload onto a GKE Autopilot cluster configured as a continuous DaemonSet pod across all nodes.
Answer
Provision Compute Engine instances utilizing Spot VMs for the batch task execution.
Spot VMs provide significant cost savings (60-91% discount) for compute workloads on Compute Engine. Because the workload is a batch process running 4 hours nightly and regularly checkpoints its progress to Cloud Storage, it can withstand VM preemption and resume without data loss.
Step-by-Step Solution
Key Concept
Planning Compute Engine Workloads using Spot VMs for Ephemeral Fault-Tolerant Tasks
Estimated Time:1m 30s