Examkin

Question

Difficulty: Very hardDesigning Infrastructure for Business Requirements and Cost Optimization

A global genomics research enterprise receives daily batch uploads totaling 50 TB50\text{ TB} of raw DNA sequencing data from partner laboratories around the world. The uploads occur unpredictably throughout the day. Each file requires a 20-minute containerized processing pipeline to extract variant metadata before storing the results. The raw files must remain instantly available for audit for 30 days, after which they are accessed less than once per year but must be retained for 7 years to meet compliance laws. The business goal is to minimize compute and storage infrastructure costs while avoiding server maintenance overhead. Which two architectural decisions should you recommend? (Select TWO)

  1. Execute the processing workload using Cloud Run jobs triggered automatically by Eventarc Cloud Storage object creation events.Answer
  2. Implement Cloud Storage Object Lifecycle Management to transition raw data from Standard to Nearline storage after 30 days, and to Archive storage after 365 days.Answer
  3. C
    Deploy a Google Kubernetes Engine (GKE) Autopilot cluster running a continuous deployment set of pods that poll Cloud Storage for new file uploads.
  4. D
    Ingest the raw sequencing files directly into Cloud Spanner instances configured with multi-region replication to provide global consistency and immediate audit retrieval.

Answer

The optimal architecture combines Cloud Run jobs triggered by Eventarc Cloud Storage events for serverless, cost-effective batch execution, alongside Cloud Storage Object Lifecycle Management to transition raw data to Nearline after 30 days and Archive after 1 year.
Combining Cloud Run jobs (triggered via Eventarc on Cloud Storage object creation) with Cloud Storage Object Lifecycle Management minimizes total cost of ownership (TCO). Cloud Run jobs scale dynamically per file execution without incurring idle compute expenses for the 20-minute processing tasks, while Lifecycle Management automatically shifts 50 TB50\text{ TB} of daily uploads into lower-cost Nearline and Archive tiers as data ages.

Step-by-Step Solution

1
Analyze compute requirements and cost profile.
The processing task runs for 20 minutes per file upon unpredictable file upload events.
Cloud Run jobs support executions up to 24 hours and scale to zero when no jobs are running, eliminating idle compute costs compared to continuous GKE deployments.
2
Analyze storage access patterns and retention rules.
Files require immediate availability for 30 days, rare access for 7 years, and long-term retention.
Transitioning objects via lifecycle rules to Nearline after 30 days (fast retrieval, lower storage cost) and to Archive after 1 year optimizes cost across the 7-year lifecycle.
3
Evaluate and eliminate sub-optimal architectural options.
Reject GKE and Cloud Spanner recommendations.
GKE introduces unnecessary cluster overhead and baseline costs for event-driven batch jobs, while Cloud Spanner is designed for relational transactional data and is extremely expensive for storing petabytes of raw un-structured files.

Key Concept

Serverless Event-Driven Batch Architecture & Storage Lifecycle Cost Optimization
Rate this question

Topics

Designing Infrastructure for Business Requirements and Cost Optimization