Optimizing Cost and Persistence for Interruptible Batch AI Workloads on Google Cloud
Short answer
B Deploy spot VMs with attached persistent disks and implement checkpoint mechanisms to balance cost, interruption tolerance, and data persistence for batch AI workloads.
Question
Key Takeaways
The main trap is ignoring the 'tolerate interruptions' clue, which points directly to Spot VMs, and failing to recognize that Persistent Disks alone aren't enough without checkpointing for state recovery.
This page explains why Spot VMs with Persistent Disks and checkpointing are the optimal solution for cost-effective, interrupt-tolerant batch AI processing on Google Cloud, matching exam PCA objectives.
Common mistake: Choosing Option A (Standard VMs) is common because it seems safer, but it fails the 'cost-effective' requirement for non-time-sensitive workloads. Choosing Option C (Local SSD) fails the 'data persistence' requirement as local storage is lost on preemption.
Expert Analysis
The question requires selecting a compute strategy for batch image processing that is cost-effective, tolerates interruptions, and ensures data persistence. The key constraints are 'accelerator-based compute,' 'not time-sensitive,' 'tolerate interruptions,' and 'data persistence.'
Spot VMs are the correct choice because they offer significant cost discounts (up to 91%) compared to standard VMs but can be preempted by Google Cloud at any time. This aligns perfectly with workloads that 'can tolerate interruptions.' Standard VMs (Option A) do not provide the same level of cost savings for interruptible workloads.
For data persistence, attaching Persistent Disks (PDs) is essential. Unlike local SSDs or ephemeral storage, PDs survive instance termination and preemption. However, since Spot VMs can be terminated unexpectedly, simply attaching a PD is not enough; one must also implement checkpoint mechanisms to save application state periodically. This ensures that if a Spot VM is preempted, the workload can resume from the last checkpoint on a new instance without losing significant progress.
Option C suggests using local SSDs, which are ephemeral and lost upon instance termination, violating the 'data persistence' requirement. Option D suggests Cloud Run functions, which are serverless and generally not suitable for long-running, accelerator-heavy batch processing tasks requiring specific GPU/TPU configurations in the same way VMs do, nor do they inherently solve the persistence issue for this specific architecture pattern described.
Therefore, deploying Spot VMs with attached Persistent Disks and implementing checkpoint mechanisms addresses all requirements: cost-effectiveness via Spot pricing, handling interruptions via preemption tolerance, and ensuring data persistence via PDs and checkpoints.
Exam Strategy
When analyzing compute options for batch jobs, first identify if the workload is fault-tolerant. If it tolerates interruptions, immediately consider Spot VMs for cost savings. Then, check for data persistence requirements; if data must survive instance death, ensure the solution includes both durable storage (like Persistent Disks) and application-level state management (like checkpoints). Never assume ephemeral storage is sufficient for persistent needs.
Frequently Asked Questions
Why not use Local SSDs for better performance in batch processing?
Can I use Preemptible VMs instead of Spot VMs?
Community Votes
74% of anonymous learners picked answer B. Votes are pick records left by other test-takers — they are not the verified answer.
Discussion 6 contributions
Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.
From public community threads 6
-
Pime13 👍 4kindly do not select any option if you are not 100% sure.
-
cachopo 👍 1 Answered BB is the correct answer because Spot VMs offer massive compute discounts and are the ideal choice for workloads that are not time-sensitive and can tolerate interruptions. Because Spot VMs can be preempted by Google Cloud at any time, attaching Persistent Disks and implementing checkpointing mechanisms ensures that your batch image processing progress is saved. If the instance is interrupted, it can spin back up and resume exactly where it left off, successfully meeting both the cost-efficiency and data persistence requirements.
-
rkumar16d 👍 1 Answered BLook at the clues in the question: Batch workloads Not time-sensitive Can tolerate interruptions → strong clue for Spot VMs Accelerator-based compute → GPUs/TPUs can be attached where supported Cost-effective → Spot VMs are the appropriate choice for interruptible workloads Data persistence when necessary → Persistent Disk Recover after interruption → checkpoint processing state periodically
-
679c828 👍 1You have posted multiple replies to incorrect answers showing intentional and malicious intent of choosing bad answers. You should be banned.
-
679c828 👍 1 Answered BAnswer B: Spot VMs for low cost and the disk check points for persistence.
Show 1 more comment
-
Gabrycloud 👍 1 Answered DI’m going with D because the others are clearly non-starters.
Reader comments & corrections
No reader corrections yet — spotted an error or have a note? Share it below.
Related Analysis
Practice All PCA Questions
Access 81 questions with complete answers and detailed explanations.
View Full PCA Practice Test →