Examkin

Question

Difficulty: MediumAuto-scaling and Capacity Planning

An enterprise supply chain platform uses worker virtual machines deployed in a Compute Engine Managed Instance Group (MIG) to process tracking updates from Cloud Pub/Sub. During peak shipment hours, message backlog accumulates significantly in Pub/Sub, but the MIG autoscaler fails to add instances because CPU utilization on existing workers remains around 25% due to external I/O wait times. Additionally, during major sales events, scaling operations occasionally halt because the project reaches regional compute limits. Which TWO actions should the Cloud Architect implement to ensure reliable auto-scaling and capacity availability? (Select TWO.)

  1. Configure the MIG autoscaler to scale based on Cloud Pub/Sub queue depth using a custom Cloud Monitoring metric calculated as unacknowledged messages per worker instance.Answer
  2. Request regional Compute Engine CPU quota increases in advance for the target deployment regions prior to expected peak traffic events.Answer
  3. C
    Lower the target CPU utilization threshold on the MIG autoscaler policy from 80% down to 15% so scaling triggers sooner during queue accumulation.
  4. D
    Migrate the processing workers to a Google Kubernetes Engine (GKE) cluster to automatically bypass regional GCP compute quota restrictions during scaling.
  5. E
    Enable dynamic regional quota auto-bursting on the Managed Instance Group to provision temporary instances beyond project quota caps during peak events.

Answer

The correct architecture requires scaling the Managed Instance Group using Cloud Pub/Sub queue depth metrics per worker instead of CPU utilization, and requesting regional Compute Engine quota increases in advance of peak events.
Scaling queue-based processing workloads effectively requires measuring queue backlog (such as Pub/Sub unacknowledged messages per worker) via Cloud Monitoring custom metrics. Additionally, because compute autoscaling cannot exceed regional quota limits, architects must plan capacity and request regional CPU quota increases prior to expected high-traffic events.

Step-by-Step Solution

1
Identify the primary scaling metric bottleneck
Recognize that CPU utilization is an ineffective scaling metric for I/O-bound pub/sub processing workers.
Workers spend time waiting on external API responses, keeping CPU low despite high queue depth.
2
Implement metric-based autoscaling
Configure MIG autoscaling based on a custom Cloud Monitoring metric reflecting Pub/Sub message backlog divided by instance count.
Queue depth accurately reflects demand and triggers scaling regardless of worker CPU load.
3
Address infrastructure capacity limits
Evaluate current regional quotas and submit quota increase requests ahead of high-demand periods.
Autoscaling policies cannot provision instances beyond the project's approved regional resource quotas.

Key Concept

Auto-scaling for I/O-bound queue workloads and proactive regional capacity planning
Estimated Time:2m 0s
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic