Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Professional Cloud Architect practice bank has 2,335 more like it. Try 10 free: instant score, no account.
An e-commerce company operates a customer-facing stateless web tier hosted on Compute Engine virtual machines across a Managed Instance Group (MIG). The application experiences predictable daily traffic surges every morning, but new VM instances take several minutes to complete initialization and warm local caches, leading to elevated response latency during reactive scale-out events. The company wants to maintain optimal performance during peak hours while minimizing compute costs during low-traffic periods.
Which autoscaling and resource optimization strategy should the company implement?
Predictive autoscaling is an advanced scaling feature in Compute Engine Managed Instance Groups (MIGs) designed to forecast future traffic patterns using machine learning models trained on historical workload metrics. Rather than only reacting to spikes as they happen, predictive autoscaling proactively provisions VM instances ahead of anticipated cyclical surges while maintaining standard autoscaling policies for unexpected demand.

Alternative approaches either require expensive static provisioning around the clock or rely purely on reactive metrics that cause response latency while instances initialize. Predictive autoscaling provides the exact balance of high performance during traffic ramp-up and cost optimization during lulls, fully adhering to the Google Cloud Well-Architected Framework.
A stateful Managed Instance Group (Stateful MIG) preserves the unique state of individual VM instances (such as persistent disks, IP addresses, and metadata) across instance restarts or recreations. Scheduled manual resizing uses an orchestrator like Cloud Scheduler to trigger API calls that adjust the target size of a group at specified times.
The workload were a stateful database or legacy application that required specific disk data and static IP configurations to be preserved across pre-planned maintenance windows.
According to the Google Cloud Well-Architected Framework, stateless workloads should leverage stateless MIGs with automated autoscaling policies to minimize cost and maximize operational agility.
A custom machine type allows administrators to tailor specific vCPU and memory allocations on Compute Engine. Disabling simultaneous multithreading (SMT) ensures each vCPU maps to a dedicated physical CPU core, which can optimize licensing costs for third-party software charged per core.
The workload were a legacy monolithic single-node application requiring specialized third-party per-core licenses that cannot run across distributed nodes.
The Google Cloud Well-Architected Framework strongly advises eliminating SPOFs by distributing stateless workloads across autoscaled Managed Instance Groups rather than relying on vertical scaling.
Spot VMs are excess Compute Engine capacity offered at steep discounts (60–91% compared to standard on-demand pricing) but are subject to preemption by Google Cloud whenever capacity is reclaimed. Reactive autoscaling adjusts the group size only after observed metrics (like average CPU utilization) cross a defined threshold.
The workload were a non-critical batch processing pipeline, rendering task, or experimental development environment that can tolerate sudden VM interruptions and delays.
While Spot VMs provide outstanding cost savings for stateless asynchronous jobs, using them for high-priority customer-facing web tiers with long initialization overhead introduces unacceptable reliability and performance risks.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated