Examkin

Question

Difficulty: MediumPlanning Compute Engine Resources and Machine Types

An online video platform is designing its Google Compute Engine compute infrastructure for two distinct backend components:

1. Media Transcoding Fleet: A fault-tolerant, stateless batch processing workload that picks up jobs from a queue and can resume interrupted tasks without data corruption.
2. Primary User Database: A continuous, high-availability relational database running 24/7 with highly predictable vCPU and memory utilization.

Which TWO deployment strategies should you select to minimize compute costs while meeting the operational availability requirements of both workloads?

  1. Provision Spot Virtual Machines (VMs) for the Media Transcoding Fleet.Answer
  2. Purchase Committed Use Discounts (CUDs) for the Primary User Database compute capacity.Answer
  3. C
    Deploy the Primary User Database on Spot VMs with auto-restart enabled to reduce licensing and infrastructure costs.
  4. D
    Apply automatic Sustained Use Discounts (SUDs) on top of Committed Use Discounts for the database instances to achieve additive savings.

Answer

Select Spot VMs for the stateless Media Transcoding Fleet to leverage steep pricing discounts for fault-tolerant jobs, and purchase Committed Use Discounts (CUDs) for the Primary User Database to secure cost savings for steady-state 24/7 compute usage.
The correct strategy leverages Spot Virtual Machines for the stateless media transcoding batch fleet because the job pipeline can easily recover from instance preemptions. Concurrently, purchasing Committed Use Discounts for the database guarantees lower pricing for steady, uninterrupted 24/7 compute consumption.

Step-by-Step Solution

1
Analyze the resiliency requirements of the Media Transcoding Fleet.
Identified as a stateless, fault-tolerant batch process capable of resuming upon interruption.
Fault-tolerant batch processing aligns directly with the operational characteristics of Spot VMs, which offer heavy discounts in exchange for potential preemption.
2
Analyze the baseline capacity requirements of the Primary User Database.
Identified as a continuous 24/7 mission-critical workload with predictable resource consumption.
Predictable 24/7 database workloads benefit most from Committed Use Discounts (CUDs), which guarantee discounted hourly compute rates for 1-year or 3-year commitments.
3
Evaluate invalid pricing and instance configuration anti-patterns.
Rejected using Spot VMs for stateful databases due to unpredictable preemption risk, and rejected combining Sustained Use Discounts with Committed Use Discounts because GCP pricing rules prevent stacking both discounts simultaneously.
Ensures architecture compliance with GCP best practices and financial estimation rules.

Key Concept

Selecting Compute Engine Purchasing Models Based on Workload Preemptibility and Usage Predictability
Rate this question

Topics

Planning Compute Engine Resources and Machine Types

More questions from this topic