Examkin

Soru

Zorluk: Çok zorPlanning Compute Engine Resources and Machine Types

An enterprise logistics company is planning the Google Cloud Compute Engine architecture for two distinct workloads:

1. An in-memory route calculation service running continuously 24/724/7 with a constant baseline demand. It requires a specific non-standard configuration of 1616 vCPUs and 128 GB128\text{ GB} of RAM (8 GB8\text{ GB} per vCPU).
2. A high-throughput image rendering pipeline that processes user-uploaded media. The rendering pipeline is stateless, fault-tolerant, and capable of resuming tasks if an instance terminates unexpectedly.

Which Compute Engine provisioning strategy optimizes infrastructure costs while meeting all operational requirements?

  1. Provision custom machine types with 16 vCPUs and 128 GB RAM covered by Committed Use Discounts (CUDs) for the route calculation service, and use Spot Virtual Machines within a Managed Instance Group for the image rendering pipeline.Cevap
  2. B
    Provision standard predefined N2 machine types covered by Committed Use Discounts for the route calculation service, and use Spot Virtual Machines for the route calculation service to lower baseline operational costs.
  3. C
    Deploy both the route calculation service and the image rendering pipeline on Cloud Functions to eliminate Compute Engine Virtual Machine management overhead entirely.
  4. D
    Provision E2 standard machine types for both workloads and rely on automatic Sustained Use Discounts (SUDs) to achieve maximum cost savings without entering contractual commitments.

Cevap

Configure custom machine types (16 vCPUs, 128 GB RAM) backed by Committed Use Discounts for the continuous route calculation service, and utilize Spot VMs in a Managed Instance Group for the stateless image rendering pipeline.
Custom machine types allow tailoring the vCPU-to-memory ratio exactly (1616 vCPUs and 128 GB128\text{ GB} RAM) to prevent paying for unused compute capacity. Because the route calculation service runs constantly (24/724/7), attaching a Committed Use Discount (CUD) maximizes savings over a 1-year or 3-year period. Conversely, the image rendering pipeline is fault-tolerant and stateless, making Spot VMs within an autoscaling Managed Instance Group the most cost-effective choice.

Adım Adım Çözüm

1
Analyze the resource ratio and execution continuity of the first workload.
The route calculation service requires 1616 vCPUs and 128 GB128\text{ GB} RAM (8 GB8\text{ GB} per vCPU), which does not align with standard predefined ratios (e.g., standard is 4 GB4\text{ GB}/vCPU, high-mem is 8 GB8\text{ GB}/vCPU on N1/N2 but fixed ratios). Using custom machine types avoids paying for unnecessary vCPUs. Since it runs 24/724/7, Committed Use Discounts (CUDs) provide up to 57% savings.
Matching custom machine specs eliminates wasted resources, and CUDs offer maximum savings for predictable, continuous compute baseline demand.
2
Evaluate fault tolerance and preemption resilience for the second workload.
The image rendering pipeline is stateless and fault-tolerant, allowing it to withstand instance terminations.
Spot VMs provide significant cost reductions (60-91% off standard prices) and are ideal for batch or stateless jobs that can handle preemption.
3
Combine the optimal pricing and machine type strategies.
Custom machine types + CUDs for Workload 1; Spot VMs in a MIG for Workload 2.
This strategy satisfies both performance/SLA guarantees and aggressive cost optimization.

Anahtar Kavram

Compute Engine Resource Planning, Custom Machine Types, Committed Use Discounts, and Spot VM Suitability
Tahmini Süre:3m 0s
Bu soruyu puanla

Konular

Planning Compute Engine Resources and Machine Types

Bu konudan daha fazla soru