A DevOps team is setting up an automated worker pool on Compute Engine to execute nightly video encoding jobs. The encoding tasks process independent chunks of data, persist progress checkpoints to Cloud Storage, and can automatically resume if an instance terminates unexpectedly. The team wants to minimize compute costs while maintaining processing capacity across the region. Which deployment strategy should the team implement?
- Create a regional Managed Instance Group utilizing Spot VM instances across multiple zones.Answer
- BProvision standard Compute Engine instances assigned with the primitive Owner IAM role to ensure compute resources are prioritized over other workloads.
- CDeploy worker instances in standard VM machine types and configure an autoscaling policy based solely on disk read throughput metrics to trigger scale-out.
- DDeploy single-zone standard VM instances and automatically launch new projects in different regions when regional CPU quotas are exceeded.
Answer
Create a regional Managed Instance Group utilizing Spot VM instances across multiple zones.
Spot VMs provide significant cost savings for batch workloads that checkpoint progress and tolerate preemption. Combining Spot VMs with a regional Managed Instance Group ensures high availability across multiple zones within the region.
Step-by-Step Solution
Key Concept
Managing Compute Engine Spot VMs and Regional Managed Instance Groups for Batch Workloads