Examkin

Question

Difficulty: MediumAuto-scaling and Capacity Planning

An enterprise financial analytics company runs an event-driven risk assessment workload on Compute Engine Managed Instance Groups (MIGs). The application receives trade messages from a queue, processes heavy calculations, and stores the results in a database. During periods of sudden market volatility, trade queue lengths increase dramatically while worker VM CPU utilization remains low due to database connection waiting times. Additionally, rapid scaling attempts during recent market events failed because the project hit regional instance limit ceilings. Which TWO architectural actions should a Cloud Architect implement to ensure auto-scaling reliability and adequate operational capacity during peak events? (Select TWO.)

  1. Configure the MIG auto-scaler to scale based on a custom Cloud Monitoring metric representing unprocessed message queue depth instead of CPU utilization.Answer
  2. B
    Configure the MIG auto-scaler using a strict 70% target average CPU utilization metric across the instance group.
  3. Request regional Compute Engine CPU quota increases in advance and implement Compute Engine capacity reservations for baseline capacity in target zones.Answer
  4. D
    Migrate the processing workers to Google Kubernetes Engine (GKE) clusters so that GKE automatically bypasses regional GCP project resource quotas.
  5. E
    Purchase 3-year term-based Committed Use Discounts (CUDs) covering the peak auto-scaling instance capacity requirements to reduce burst scaling latency.

Answer

The correct recommendations are configuring MIG auto-scaling using custom queue depth metrics in Cloud Monitoring, and requesting regional CPU quota increases in advance alongside Compute Engine capacity reservations.
Scaling based on custom metrics (such as queue depth) ensures I/O-bound workers scale according to true backlog demand rather than low CPU utilization. Additionally, requesting proactive quota increases and creating capacity reservations ensures that both logical quota room and physical infrastructure are guaranteed during unexpected traffic spikes.

Step-by-Step Solution

1
Analyze workload scaling triggers
Identify that the workload bottleneck is queue backlogs and I/O wait times rather than CPU saturation.
CPU-based scaling fails for I/O-bound worker pools; Cloud Monitoring custom queue depth metrics trigger scaling proportional to incoming trade message backlogs.
2
Analyze capacity management and quota constraints
Identify that auto-scaling cannot exceed regional GCP quotas or guarantee hardware availability without explicit configuration.
Requesting regional quota increases prevents API provision errors during rapid scale-up, while zonal capacity reservations ensure physical Compute Engine resources are reserved for peak events.

Key Concept

Custom Metric Auto-scaling & Compute Capacity Planning
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic