Examkin

Question

Difficulty: MediumAuto-scaling and Capacity Planning

An enterprise e-commerce platform processes asynchronous product image re-indexing and media processing jobs using Compute Engine Managed Instance Groups (MIGs) fed by a Cloud Pub/Sub topic. During high-traffic marketing events, the worker nodes experience network I/O wait state spikes, resulting in processing backlogs even though average CPU utilization remains below 35%. Additionally, rapidly launching new instances during traffic surges occasionally fails due to resource availability boundaries. Which TWO architectural actions should you implement to ensure effective auto-scaling and capacity reliability? (Select TWO)

  1. Configure the MIG autoscaler to scale based on a custom Cloud Monitoring metric tracking Pub/Sub unacknowledged message backlog per instance rather than CPU utilization.Answer
  2. B
    Set the target CPU utilization threshold on the MIG autoscaler to 90% to force instances to trigger scaling when I/O processing delays accumulate.
  3. Request regional Compute Engine quota increases for vCPUs and in-use IP addresses well before expected high-traffic events to prevent capacity allocation failures.Answer
  4. D
    Migrate the worker nodes to Google Kubernetes Engine (GKE) so that Cluster Autoscaler automatically bypasses regional GCP project resource quotas.
  5. E
    Replace Compute Engine MIGs with stateless GKE clusters to avoid setting up custom metrics for I/O-bound queue processing workloads.

Answer

The correct actions are to configure the MIG autoscaler to scale based on custom metrics such as Pub/Sub unacknowledged message backlog per instance, and to proactively request regional Compute Engine quota increases for vCPUs and IP addresses prior to high-traffic events.
Scaling queue-based, I/O-bound processing requires metrics that track queue depth or backlog per instance because CPU utilization stays low during network and storage waits. Furthermore, auto-scaling relies on underlying GCP infrastructure capacity; requesting regional vCPU and networking quota increases in advance guarantees that instances can launch successfully when scale-out conditions are met.

Step-by-Step Solution

1
Identify the workload metric bottleneck
Recognize that I/O-bound asynchronous queue processing workloads keep CPU utilization artificially low while work accumulates.
Scaling on CPU utilization fails for I/O-bound tasks; tracking queue depth or custom metrics accurately reflects scaling demand.
2
Identify capacity planning requirements
Recognize that auto-scaling mechanisms cannot exceed pre-allocated GCP project resource quotas.
Proactive quota management prevents scaling execution failures when traffic surges require launching additional instances.

Key Concept

Auto-scaling metric selection for I/O-bound workloads and proactive regional quota management
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic