An online education platform is designing a new Google Kubernetes Engine (GKE) cluster architecture to handle on-demand video encoding tasks. The video encoding workloads are stateless, fault-tolerant, and can tolerate sudden instance preemptions. To reduce operational burden, the engineering leadership has mandated that node management overhead—including node provisioning, security patching, and OS upgrades—must be completely managed by Google Cloud. Which cluster design strategy fulfills all technical requirements while minimizing compute expenses?
- Deploy a GKE Autopilot cluster and configure the video encoding workloads to request Spot Pods using node selectors or tolerations.Answer
- BProvision a GKE Standard cluster with node pools composed of Compute Engine Spot VMs and enable Cluster Autoscaler.
- CDeploy a GKE Autopilot cluster and rely on Horizontal Pod Autoscaler to adjust the compute node count based on CPU utilization.
- DProvision a GKE Standard cluster with a single zonal node pool of standard On-Demand VM instances and enable node auto-repair.
Answer
Deploying a GKE Autopilot cluster with workloads configured to request Spot Pods satisfies the requirement of zero node management overhead while achieving maximum cost efficiency for fault-tolerant tasks.
GKE Autopilot shifts node management responsibilities completely to Google Cloud, satisfying the requirement to eliminate node maintenance overhead. By configuring Kubernetes manifests to target Spot Pods (via tolerations and node selectors), the cluster runs stateless video encoding tasks on discounted compute capacity without requiring manually configured Spot VM node pools.
Step-by-Step Solution
Key Concept
GKE Autopilot Cluster Architecture and Spot Pod Integration
Estimated Time:1m 30s