Examkin

Question

Difficulty: MediumSelecting and Designing Compute and Application Platforms

A financial analytics company is designing a new cloud architecture on Google Cloud for two distinct backend services within their risk analysis pipeline.

1. Risk Ingestion Service: A high-throughput, stateless HTTP/2 REST API microservice that receives incoming market data webhooks. Traffic fluctuates heavily throughout the day, dropping to zero during market off-hours. It requires zero server maintenance, fast cold-starts, and scaling down to zero instances to minimize idle costs.
2. Monte Carlo Simulation Engine: A high-performance compute workload requiring custom Linux kernel modules, direct hardware access to NVIDIA GPUs, local NVMe SSD scratch storage for fast state caching, and long-running batch job execution lasting up to 14 hours.

Which TWO platform selection decisions should you recommend to fulfill these workload requirements while optimizing cost and operational efficiency? (Select TWO.)

  1. Deploy the Risk Ingestion Service on Cloud Run to take advantage of automatic scaling to zero, pay-per-use billing, and zero cluster management overhead.Answer
  2. B
    Provision a multi-zone Google Kubernetes Engine (GKE) cluster with Standard mode node pools to host the Risk Ingestion Service to ensure minimal latency during traffic spikes.
  3. Deploy the Monte Carlo Simulation Engine on Compute Engine Managed Instance Groups (MIGs) utilizing custom VM images with required kernel modules, GPU attachments, and local SSDs.Answer
  4. D
    Deploy the Monte Carlo Simulation Engine on App Engine Standard environment using background instances to leverage automated infrastructure scaling.
  5. E
    Purchase 3-year Committed Use Discounts (CUDs) for maximum compute instances on the Risk Ingestion Service to minimize per-request baseline costs.

Answer

Deploy the stateless HTTP Risk Ingestion Service on Cloud Run for scale-to-zero efficiency and operational simplicity, and host the Monte Carlo Simulation Engine on Compute Engine Managed Instance Groups (MIGs) to support custom kernel modules, attached GPUs, local NVMe SSDs, and 14-hour batch runs.
The solution correctly pairs the stateless HTTP API service with Cloud Run to take advantage of scale-to-zero autoscaling, zero management overhead, and request-based pricing, while assigning the complex compute engine with custom kernel, GPU, local SSD, and extended execution time requirements to Compute Engine MIGs.

Step-by-Step Solution

1
Analyze requirements for the Risk Ingestion Service
Identified a stateless, HTTP/2 REST API with variable traffic dropping to zero.
Cloud Run provides serverless container execution that automatically scales instances to zero during off-hours and handles container deployment without cluster management overhead.
2
Analyze requirements for the Monte Carlo Simulation Engine
Identified requirements for custom kernel modules, GPU acceleration, local NVMe SSDs, and long-running execution (14 hours).
Compute Engine VM instances allow full root access to modify kernel drivers, attach local NVMe storage and GPUs directly, and run unbounded compute jobs.
3
Evaluate invalid compute platform combinations
Eliminated GKE standard clusters for simple stateless ingestion due to idle baseline costs/management overhead, App Engine Standard due to hardware/kernel constraints, and fixed long-term CUD commitments for bursty zero-idle services.
Matching compute workloads strictly to architectural capabilities avoids misallocating infrastructure.

Key Concept

GCP Compute Platform Selection based on statefulness, execution duration, hardware dependencies, and operational overhead limits.
Estimated Time:2m 0s
Rate this question

Topics

Selecting and Designing Compute and Application Platforms