Examkin

Question

Difficulty: Very hardPlanning Compute Engine Resources and Machine Types

An enterprise financial institution is designing the Compute Engine infrastructure for two core workloads:
1. A mission-critical, stateful transaction processing service that operates continuously 24/7 with a predictable high-memory footprint and zero tolerance for unexpected instance terminations.
2. An overnight Monte Carlo risk simulation batch engine that executes thousands of independent, stateless calculation tasks with native checkpointing.

Which TWO provisioning and cost-optimization strategies should you recommend to minimize operational costs while satisfying all SLA requirements? (Select TWO answers.)

  1. Purchase 3-year Committed Use Discounts (CUDs) for the baseline compute capacity required by the stateful transaction processing service.Answer
  2. Provision the Monte Carlo risk simulation batch engine using Spot Virtual Machines across multiple availability zones.Answer
  3. C
    Provision the stateful transaction processing service on Spot VMs configured with custom memory-optimized machine types.
  4. D
    Rely on automatic Sustained Use Discounts (SUDs) on standard compute instances for the Monte Carlo risk simulation batch engine rather than using Spot VMs.

Answer

The optimal architecture combines purchasing 3-year Committed Use Discounts (CUDs) for the 24/7 stateful transaction processing service and provisioning the Monte Carlo risk simulation batch engine using Spot Virtual Machines.
Purchasing Committed Use Discounts (CUDs) provides deep savings for baseline 24/7 stateful workloads without risking node preemption, while Spot VMs provide maximum cost reduction (60-91%) for fault-tolerant overnight batch jobs that support native checkpointing.

Step-by-Step Solution

1
Analyze workload characteristics and SLA requirements for the transaction processing service.
The service is 24/7, stateful, has a predictable compute/memory profile, and cannot tolerate unexpected preemption.
Steady-state uninterruptible workloads qualify for Committed Use Discounts (CUDs), yielding up to 70% cost reduction without operational risk.
2
Analyze workload characteristics and fault-tolerance for the Monte Carlo simulation batch engine.
The engine is stateless, runs periodically/overnight, supports checkpointing, and tolerates node interruptions.
Fault-tolerant batch jobs with checkpointing are ideal candidates for Spot VMs, which offer deep discounts of 60% to 91% compared to standard pricing.
3
Evaluate and eliminate non-optimal or risky compute options.
Using Spot VMs for stateful zero-downtime services risks application failure, while using standard VMs with Sustained Use Discounts for intermittent batch jobs yields suboptimal savings.
Spot VMs can be revoked with a 30-second notice, and Sustained Use Discounts require prolonged continuous execution on standard instances.

Key Concept

Selecting Compute Engine purchasing and provisioning models based on workload persistence and availability SLAs
Rate this question

Topics

Planning Compute Engine Resources and Machine Types

More questions from this topic