An enterprise runs three distinct workloads on Google Cloud Compute Engine: Core Backend Service: A mission-critical, steady-state service running 24/7. Cloud Monitoring utilization metrics show continuous consumption of exactly 24 vCPUs and 48 GB RAM (a 1:2 CPU-to-memory ratio). Web Frontend Tier: A stateless application tier experiencing diurnal, traffic-driven scaling fluctuations. Image Processing Pipeline: An asynchronous, fault-tolerant batch worker fleet that can tolerate sudden task interruptions. Which combination of instance configurations and purchasing models should the DevOps engineer implement to minimize total infrastructure costs while maintaining reliability?

Provision custom machine types (24 vCPUs, 48 GB RAM) covered by resource-based Committed Use Discounts (CUDs) for the backend, deploy a stateless Managed Instance Group (MIG) with autoscaling for the web frontend, and use Spot VMs for the batch processing pipeline.: ### What the combined optimization strategy is This strategy combines **custom machine types**, **resource-based Committed Use Discounts (CUDs)**, **stateless Managed Instance Groups (MIGs) with autoscaling**, and **Spot VMs** to align infrastructure purchase models precisely with workload profiles and utilization metrics. ### How it meets the requirements - **Core Backend Service (Right-sizing & CUDs)**: Predefined general-purpose machine types (such as `n1-standard` or `n2-standard`) enforce standard resource ratios of 1 vCPU to 4 GB RAM. For a workload requiring 24 vCPUs and 48 GB RAM (1:2 ratio), predefined sizing would force provisioning a standard 24 vCPU / 96 GB RAM instance or an oversized machine, creating 48 GB of unutilized, paid RAM. Provisioning a **custom machine type** aligns allocation directly to the 1:2 ratio. Backing this steady-state baseline with a 1-year or 3-year **resource-based Committed Use Discount** cuts baseline compute costs by up to 55% without incurring waste. - **Web Frontend Tier (Autoscaling MIGs)**: Stateless MIGs with autoscaling dynamically adjust compute capacity to traffic volume, eliminating idle capacity during off-peak hours. - **Image Processing Pipeline (Spot VMs)**: Spot VMs offer a 60% to 91% discount on vCPUs and memory for fault-tolerant batch jobs that can handle preemption without affecting core availability. ### Key advantages and capabilities - **Zero Resource Waste**: Tailors memory and CPU specifications to empirical Cloud Monitoring data. - **Significant Baseline Discounts**: Applies CUDs to predictable 24/7 baseline demand. - **Deep Spot Savings**: Reduces batch processing spend by up to 91% using excess cloud capacity. - **Elastic Scaling**: Dynamically handles diurnal web traffic bursts using managed autoscaling. ### Why this is optimal This architecture adheres to core FinOps principles by pairing predictable, non-standard footprints with custom sizing and commitments, while utilizing elasticity and deeply discounted preemptible capacity where architectural fault tolerance permits.

Compute Engine Right Sizing Cost Models | Professional Cloud DevOps Engineer

Compute Engine Right Sizing Cost Models

Worked example. The correct answer is already marked and every option is explained below, so there is nothing to select here. To answer questions yourself, start the free trial.

An enterprise runs three distinct workloads on Google Cloud Compute Engine:

Core Backend Service: A mission-critical, steady-state service running 24/7. Cloud Monitoring utilization metrics show continuous consumption of exactly 24 vCPUs and 48 GB RAM (a 1:2 CPU-to-memory ratio).
Web Frontend Tier: A stateless application tier experiencing diurnal, traffic-driven scaling fluctuations.
Image Processing Pipeline: An asynchronous, fault-tolerant batch worker fleet that can tolerate sudden task interruptions.

Which combination of instance configurations and purchasing models should the DevOps engineer implement to minimize total infrastructure costs while maintaining reliability?

Deploy the backend service and the batch processing pipeline on Spot VMs within a regional Managed Instance Group with autohealing, and purchase resource-based CUDs for peak autoscaled capacity on the web frontend.

Provision custom machine types (24 vCPUs, 48 GB RAM) covered by resource-based Committed Use Discounts (CUDs) for the backend, deploy a stateless Managed Instance Group (MIG) with autoscaling for the web frontend, and use Spot VMs for the batch processing pipeline.

Provision standard predefined memory-optimized machine types (m1 series) covered by 3-year spend-based CUDs across all tiers, including the batch image processing pipeline.

Standardize all instances on predefined general-purpose n2-standard-32 machine types, apply automatic sustained use discounts (SUDs), and disable horizontal autoscaling to avoid operational overhead.

Professional Cloud DevOps Engineer

Compute Engine Right Sizing Cost Models

Related questions