Examkin

Soru

Zorluk: OrtaSelecting and Designing Compute and Application Platforms

A financial enterprise is re-architecting its nightly batch risk calculation system on Google Cloud. The system executes containerized, stateless tasks triggered on a fixed schedule. Each batch task requires up to 45 minutes of processing time, requires custom CPU and memory allocations, and must scale to zero when no jobs are running to avoid idle infrastructure costs. Which TWO architectural decisions should the team make to meet these requirements with minimal operational overhead?

  1. Execute the batch workloads using Cloud Run jobs, configuring the execution timeout to accommodate the 45-minute task duration.Cevap
  2. B
    Provision a dedicated GKE standard cluster with static node pools to host the scheduled batch containers.
  3. Configure Cloud Scheduler to trigger the Cloud Run job executions directly using service account authentication.Cevap
  4. D
    Purchase 3-year Committed Use Discounts (CUDs) on Compute Engine VM instances reserved exclusively for running these nightly jobs.
  5. E
    Deploy the containers to a Compute Engine Managed Instance Group (MIG) configured with autoscaling based exclusively on CPU utilization metrics.

Cevap

The team should execute the batch workloads using Cloud Run jobs with appropriate execution timeouts and trigger them via Cloud Scheduler with service account authentication.
Cloud Run jobs are purpose-built for containerized tasks that run to completion without serving incoming web requests. They support execution times up to 24 hours and charge only for resources consumed during execution, automatically scaling down to zero when idle. Paired with Cloud Scheduler, the entire solution operates seamlessly without requiring persistent server infrastructure, minimizing both operational overhead and costs.

Adım Adım Çözüm

1
Analyze the workload characteristics and operational constraints.
Workload is stateless, containerized, scheduled, requires up to 45 minutes of run time, and needs to scale to zero to minimize idle costs.
Matching workload execution patterns to the appropriate GCP compute platform ensures cost efficiency and low management overhead.
2
Evaluate compute platforms against task execution duration and cost requirements.
Cloud Run jobs support tasks running up to 24 hours and scale to zero completely when idle, avoiding GKE cluster overhead or always-on VM costs.
Cloud Run services cap HTTP request timeouts at 60 minutes, but Cloud Run jobs are designed specifically for long-running batch tasks.
3
Select the serverless orchestration mechanism for triggering scheduled executions.
Cloud Scheduler provides a fully managed serverless cron service that securely triggers Cloud Run jobs via IAM authentication.
Combining serverless scheduling with serverless execution provides an end-to-end architecture with zero persistent infrastructure management.

Anahtar Kavram

Selecting serverless compute platforms for scheduled long-running batch workloads
Bu soruyu puanla

Konular

Selecting and Designing Compute and Application Platforms