Examkin

Question

Difficulty: HardCapacity Planning and Infrastructure Workload Scaling Optimization

A global logistics enterprise runs a real-time inventory tracking platform on Google Cloud. Telemetry data from IoT sensors is ingested via Cloud Pub/Sub and processed by memory-intensive worker applications running on Compute Engine Managed Instance Groups (MIGs). During simulation testing for an upcoming peak event, Pub/Sub message queues backed up rapidly because the MIG autoscaler failed to add instances quickly enough under heavy memory pressure when configured with default CPU utilization metrics. Furthermore, instance creation attempts stalled mid-scaling due to reaching regional vCPU limits. Which TWO actions should you take to optimize workload scaling and capacity planning for this platform? (Select TWO.)

  1. Configure the Managed Instance Group autoscaling policy using custom Cloud Monitoring metrics tracking Pub/Sub unacknowledged message age and worker memory utilization.Answer
  2. B
    Maintain CPU utilization as the primary autoscaling metric for the MIGs and migrate the underlying database from Cloud SQL to Cloud Spanner to absorb queue backlog pressure.
  3. Audit existing resource quotas in the target region and submit requests for vCPU quota increases well in advance of the anticipated peak load.Answer
  4. D
    Migrate the processing workers to a Google Kubernetes Engine (GKE) cluster and rely on default GKE Cluster Autoscaler settings to handle compute expansion.

Answer

The platform requires configuring custom Cloud Monitoring metrics (such as Pub/Sub unacknowledged message age and memory utilization) for MIG autoscaling, along with proactively requesting regional vCPU quota increases prior to peak workload events.
To optimize workload scaling for memory-intensive streaming workers processing Pub/Sub queues, autoscaling must be driven by custom metrics reflecting queue backlog (unacknowledged message age) and worker memory consumption. Additionally, capacity planning demands auditing regional GCP quotas and requesting vCPU quota increases ahead of time to avoid provisioning failures during peak demand spikes.

Step-by-Step Solution

1
Analyze the autoscaling bottleneck for the memory-intensive and queue-bound workload.
Default CPU utilization metrics fail to trigger autoscaling when workers are bound by memory saturation or growing Pub/Sub message queues.
Autoscaling policies must align with the actual resource bottleneck (queue depth and memory) using custom metrics.
2
Identify capacity limit constraints during burst scaling events.
Provisioning attempts failed due to reaching regional Compute Engine quotas during peak traffic spikes.
Capacity planning requires reviewing regional quotas and requesting quota increases in advance of anticipated scale-out events.

Key Concept

Workload Autoscaling Optimization and Regional Capacity Planning
Rate this question

Topics

Capacity Planning and Infrastructure Workload Scaling Optimization