Examkin

Question

Difficulty: MediumCapacity Planning and Infrastructure Workload Scaling Optimization

An enterprise fleet management company processes batch vehicle telemetry and sensor logs uploaded by 50,000 connected delivery vans at the end of each shift. The processing architecture uses Compute Engine Managed Instance Groups (MIGs) consuming messages from a Cloud Pub/Sub topic. During peak shift handovers, message processing backlogs increase rapidly, but worker VMs fail to scale out quickly enough because autoscaling is triggered only by CPU load. Furthermore, sudden cluster scaling periodically fails due to reaching regional vCPU quota limits in us-central1. Which TWO actions should the Cloud Architect take to resolve the processing backlog and ensure reliable infrastructure scaling capacity? (Select TWO.)

  1. Configure the MIG autoscaler to use a custom metric tracking the number of unacknowledged Pub/Sub messages per worker instance.Answer
  2. Perform proactive capacity planning by auditing workload growth and submitting regional vCPU quota increase requests prior to high-volume events.Answer
  3. C
    Replace the Compute Engine MIG workers with a Google Kubernetes Engine (GKE) cluster for basic queue consumer scripts to eliminate the need for capacity planning.
  4. D
    Rely on automatic runtime scaling to auto-expand regional vCPU quotas dynamically when worker instances hit provisioning limits during peak loads.

Answer

The Cloud Architect should configure the MIG autoscaler to scale based on Pub/Sub queue depth custom metrics per worker instance and proactively submit regional vCPU quota increase requests in advance of peak load periods.
For message-driven consumer workloads on Google Cloud Compute Engine, autoscaling should be tied to custom metrics such as Pub/Sub unacknowledged message depth per instance rather than CPU utilization. Additionally, to avoid VM creation failures during rapid autoscaling events, cloud architects must perform capacity planning and request regional vCPU quota increases in advance.

Step-by-Step Solution

1
Analyze workload characteristics and metric bottlenecks.
Identified that CPU utilization is an inadequate autoscaling metric for queue-based batch consumer workloads.
Queue workers may be network or I/O bound while waiting for message batches, causing CPU-based autoscalers to lag behind growing message backlogs.
2
Select the appropriate scaling signal.
Configured Pub/Sub unacknowledged message count per instance as the custom metric for the MIG autoscaler.
Tracking queue depth directly aligns worker VM provisioning with actual backlog size.
3
Address compute capacity limits.
Forecast peak compute requirements and request regional vCPU quota increases prior to deployment bursts.
Proactive quota management prevents instance provisioning failures when autoscaling triggers rapid compute expansion.

Key Concept

Workload autoscaling using custom queue metrics and proactive regional capacity quota planning.
Rate this question

Topics

Capacity Planning and Infrastructure Workload Scaling Optimization