Examkin

Question

Difficulty: HardPlanning Compute Engine Resources and Machine Types

An enterprise analytics team is designing the infrastructure for a memory-intensive transactional database on Google Compute Engine. Benchmarking indicates the database requires 416 GB416\text{ GB} of RAM and 32 vCPUs32\text{ vCPUs}. The database is strictly stateful, operates 24/7 with zero tolerance for ungraceful instance terminations, and is projected to run continuously for at least three years. Which machine family and purchasing model configuration should the cloud engineer select to meet these operational and financial requirements?

  1. Provision an m1-ultramem-32 memory-optimized Compute Engine instance attached to a 3-year Committed Use Discount (CUD).Answer
  2. B
    Provision a Spot VM instance using a custom machine type configured with 32 vCPUs and 416 GB of RAM to minimize hourly compute costs.
  3. C
    Deploy the database workload to GKE Autopilot using custom pod resource requests of 32 vCPUs and 416 GB of RAM to leverage automated node management.
  4. D
    Deploy the database container onto Cloud Run configured with maximum vCPU allocation and an attached network storage volume.

Answer

Provisioning an m1-ultramem-32 memory-optimized Compute Engine instance with a 3-year Committed Use Discount (CUD) is the correct choice.
The Memory-Optimized machine family (M1/M2/M3) is engineered specifically for memory-intensive workloads such as large in-memory databases like SAP HANA or Redis clusters requiring high RAM-to-vCPU ratios. Combined with a 3-year Committed Use Discount (CUD), this configuration guarantees dedicated baseline availability without preemption risks while achieving maximum cost efficiency for a continuous multi-year deployment.

Step-by-Step Solution

1
Analyze workload resource requirements and select machine family.
The requirement calls for 416 GB RAM with 32 vCPUs (a ratio of 13 GB RAM per vCPU). Memory-optimized families (M1/M2/M3) offer up to 28 GB RAM per vCPU, ideal for large memory-bound databases.
General-purpose or compute-optimized machine types do not provide adequate memory density for 416 GB RAM without over-provisioning CPU cores.
2
Evaluate fault tolerance and availability constraints.
The workload is stateful and requires zero ungraceful terminations, ruling out Spot/Preemptible VMs.
Spot instances can be revoked at any time when compute capacity is needed elsewhere.
3
Determine the optimal pricing strategy for a 3-year continuous runtime.
Select a 3-year Committed Use Discount (CUD) for Compute Engine vCPUs and memory.
Committed Use Discounts offer the highest savings (up to 57% or 70% depending on resource type) for baseline, steady-state workloads running continuously over 1 or 3 years.

Key Concept

Selecting Compute Engine Machine Types and Purchasing Models for Memory-Bound Workloads
Rate this question

Topics

Planning Compute Engine Resources and Machine Types

More questions from this topic