Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Associate Cloud Engineer practice bank has an additional 2,014 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A company is designing a compute architecture for stateless batch-processing and data analysis workloads on Google Cloud. These workloads are fault-tolerant, do not require continuous uptime, and can be safely stopped and resumed if infrastructure resources are reclaimed.
The team wants to minimize compute costs as much as possible while maintaining the ability to launch compute instances on demand without reserving capacity in advance.
Which instance provisioning model should the company configure for these compute instances?
The Flex-start provisioning model is designed for use with Dynamic Workload Scheduler (DWS) to run specialized batch AI and machine learning workloads. It provides specific discounts when provisioning supported accelerator-optimized machine series (such as A3, A4, or H4D).
The workload specifically required provisioning GPU-heavy AI/ML training clusters on supported machine families (e.g., A3 or A2 instances) using the Dynamic Workload Scheduler.
For standard general-purpose batch processing and analysis, Spot VMs provide broader machine type compatibility and higher cost efficiency without hardware-specific scheduler constraints.
Spot instances (and Spot VMs) are compute instances that run on surplus Google Cloud capacity. They provide the exact same machine types, capabilities, and performance as standard compute instances but at a substantial discount compared to standard on-demand pricing. In exchange for the discounted pricing, Google Cloud can preempt (reclaim) these instances whenever the underlying capacity is required for standard workloads.
For stateless, fault-tolerant batch workloads, Spot instances deliver optimal economic efficiency without requiring long-term financial commitments or upfront reservation planning.
The Reservation-bound provisioning model requires capacity to be reserved in advance through mechanisms such as future reservations in AI Hypercomputer or calendar mode future reservations. The instances can only be created once the reservation delivery date arrives.
The organization were planning a large-scale, time-sensitive training or high-performance computing job months in advance and needed guaranteed capacity delivery at a specific date and time.
Reservation-bound models are intended for predictable, high-assurance capacity needs rather than flexible, opportunistic batch jobs.
The Standard provisioning model provides regular, on-demand compute instances that run continuously until explicitly stopped or deleted by the user. Standard instances are protected by Google Cloud's availability policies, including live migration during host maintenance events.
The workload were mission-critical, stateful, or required high availability and uninterrupted execution with strict SLAs where preemption cannot be tolerated.
While Standard instances have no creation prerequisites and can launch on demand, using them for interruptible batch workloads results in significantly higher operational expenditures compared to Spot instances.
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