Examkin

Question

Difficulty: EasyPlanning Compute Engine Resources and Machine Types

An organization is planning to deploy batch machine learning training jobs on Google Cloud Compute Engine. The jobs support native checkpointing, can tolerate unexpected instance preemptions without losing progress, and must be executed at the lowest possible cost. Which Compute Engine provisioning model should you recommend?

  1. Spot VM instancesAnswer
  2. B
    Standard VM instances backed by 3-year Committed Use Discounts
  3. C
    Google Kubernetes Engine Autopilot nodes
  4. D
    Cloud Functions event-driven instances

Answer

Spot VM instances should be selected to run fault-tolerant, interruptible batch workloads at the lowest possible cost.
Spot VM instances utilize excess Compute Engine capacity and offer discounts between 60% and 91% compared to standard VM pricing. Because the machine learning workload natively checkpoints progress and tolerates preemptions, Spot VMs deliver maximum cost efficiency without risking data loss.

Step-by-Step Solution

1
Analyze the operational requirements of the workload
The machine learning training job is a batch process, supports checkpointing, and can handle unexpected interruptions.
Fault-tolerant workloads do not require guaranteed VM availability or high-availability SLAs.
2
Evaluate Google Cloud Compute Engine cost-optimization features
Spot VMs provide steep discounts by using spare Google Compute Engine capacity, with the caveat that Compute Engine can reclaim them at any time.
Because the workload saves its state periodically, preemption simply delays execution slightly rather than causing data loss, achieving maximum cost savings.

Key Concept

Planning Compute Engine Resources and Machine Types for Fault-Tolerant Workloads
Rate this question

Topics

Planning Compute Engine Resources and Machine Types

More questions from this topic