Examkin

Question

Difficulty: MediumAuto-scaling and Capacity Planning

An enterprise financial reporting service running on Compute Engine Managed Instance Groups (MIGs) processes incoming asynchronous HTTP webhooks. During market open, the service experiences severe latency and request timeouts due to queue backlogs, despite average CPU utilization staying below 25% because the workload is heavily I/O and memory bound. Furthermore, during scheduled quarterly filing events, rapid auto-scaling attempts fail because the target region runs out of available instance quotas. Which combination of actions should the Cloud Architect recommend to resolve both the auto-scaling and capacity issues?

  1. Configure the MIG autoscaler to scale using a Cloud Monitoring custom metric based on pending queue backlog, and proactively submit a regional Compute Engine quota increase request prior to quarterly filing events.Answer
  2. B
    Lower the target CPU utilization threshold on the MIG autoscaler to 15% to force scaling during queue backlogs, and submit a regional quota increase request prior to quarterly filing events.
  3. C
    Configure the MIG autoscaler to scale using a Cloud Monitoring custom metric based on pending queue backlog, while relying on Google Cloud's automatic dynamic quota allocation during traffic spikes.
  4. D
    Migrate the application to a multi-zonal Google Kubernetes Engine (GKE) cluster managed by Horizontal Pod Autoscaler (HPA) using CPU utilization metrics to bypass Compute Engine regional quota limits.

Answer

Configure the MIG autoscaler to scale using a Cloud Monitoring custom metric based on pending queue backlog, and proactively submit a regional Compute Engine quota increase request prior to quarterly filing events.
For memory and I/O-bound applications, standard CPU utilization metrics fail to reflect workload pressure. Using Cloud Monitoring custom metrics (such as queue backlog depth) allows the MIG autoscaler to respond accurately. Additionally, GCP regional quotas are fixed policy limits that do not expand automatically; submitting quota increase requests prior to planned demand spikes prevents scaling errors.

Step-by-Step Solution

1
Identify the root cause of autoscaling failure during normal spikes.
Because the workload is I/O and memory bound, CPU metrics remain low despite high queue depth. Scaling must be tied to custom Cloud Monitoring metrics such as queue length.
Standard CPU metrics do not accurately reflect capacity constraints for memory/I/O-bound workloads.
2
Identify the root cause of scaling failure during scheduled quarterly events.
MIG autoscaling cannot exceed project/regional quota limits. Submitting a quota increase request in advance ensures sufficient headroom.
Compute limits are strictly enforced by GCP IAM and quota controls and do not auto-increase during traffic surges.

Key Concept

Auto-scaling based on custom Cloud Monitoring metrics and capacity quota management
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic