Examkin

Question

Difficulty: MediumSelecting and Designing Compute and Application Platforms

A scientific research institute is migrating a specialized numerical simulation workload to Google Cloud. The simulation software requires custom Linux kernel modules that must be loaded directly into the host operating system kernel, executes continuous batch runs lasting up to 10 hours per job, and requires local high-speed NVMe storage. The operations team wants to ensure minimal cost when no simulation jobs are running while keeping infrastructure management simple. Which compute platform should the cloud architect select to meet these technical constraints?

  1. Compute Engine virtual machines utilizing custom OS images and Spot VMs, configured to terminate upon job completion.Answer
  2. B
    Google Kubernetes Engine (GKE) Autopilot mode deployed with custom DaemonSets to load the required kernel drivers.
  3. C
    Cloud Run jobs configured with max instances and Cloud Storage FUSE volume mounts.
  4. D
    Cloud Functions (2nd gen) triggered by Cloud Pub/Sub events with maximum memory allocations.

Answer

Compute Engine virtual machines utilizing custom OS images and Spot VMs, configured to terminate upon job completion.
Compute Engine is the only option that offers complete administrative access to the guest operating system, allowing the installation of custom Linux kernel modules. Additionally, Compute Engine supports jobs running up to 10 hours without timeout limits, allows local NVMe SSD attachments, and supports Spot VMs to dramatically lower compute costs for batch processing.

Step-by-Step Solution

1
Analyze OS Kernel and Infrastructure Control Requirements
The requirement for custom Linux host kernel modules eliminates fully managed serverless platforms (Cloud Run, Cloud Functions) and managed container node platforms (GKE Autopilot) where host kernel access is restricted.
Serverless and managed container platforms hide host node kernel management to maintain security and multi-tenancy abstraction.
2
Evaluate Job Duration and Storage Constraints
The 10-hour execution duration exceeds standard serverless execution limits (such as Cloud Functions' 60-minute cap), requiring IaaS Virtual Machines.
Compute Engine instances have no job duration execution caps and support direct attachment of local NVMe SSDs.
3
Optimize Cost Efficiency for Batch Workloads
Using Spot VMs with custom images allows for up to 60-91% cost savings for fault-tolerant or scheduled batch jobs.
Terminating instances when no rendering or simulation jobs are active prevents idle resource charges.

Key Concept

Selecting GCP Compute Platforms based on Kernel Control, Execution Limits, and Custom Hardware Needs
Rate this question

Topics

Selecting and Designing Compute and Application Platforms