Examkin

Question

Difficulty: HardAuto-scaling and Capacity Planning

An enterprise financial compliance application deployed on Compute Engine Managed Instance Groups (MIGs) processes continuous streaming network traffic. During high-throughput compliance audit cycles, backend instances experience severe latency and drop connections due to socket allocation exhaustion and memory pressure, while average CPU utilization remains at approximately 35%. Furthermore, upcoming quarterly compliance stress testing requires expanding the infrastructure by 300 instances, but past automated scale-out events failed immediately because the target region exceeded baseline vCPU limits. Which combination of autoscaling policy and capacity planning strategies should the Cloud Architect implement to ensure operational reliability?

  1. Configure the MIG autoscaler to use a custom Cloud Monitoring metric reflecting active open socket connections, submit a regional vCPU quota increase request well before the audit cycle, and create compute capacity reservations.Answer
  2. B
    Set the target CPU utilization of the MIG autoscaler to 30% to force earlier instance creation, and rely on standard dynamic instance scaling without reserving capacity in advance.
  3. C
    Configure custom Cloud Monitoring metric scaling based on active network sockets, and deploy an automated Cloud Function to request regional vCPU quota increases dynamically when instance creation errors occur during the stress test.
  4. D
    Migrate the application to Google Kubernetes Engine (GKE) using Horizontal Pod Autoscaling (HPA) targeting 60% CPU utilization, and deploy a custom node pool across multiple zones to bypass Compute Engine regional quota boundaries.

Answer

Configure the MIG autoscaler using a custom Cloud Monitoring metric for active socket connections, request regional vCPU quota increases in advance, and create compute capacity reservations.
The workload is limited by open socket connections and memory pressure rather than CPU utilization. Therefore, autoscaling must be configured against a custom Cloud Monitoring metric that measures socket exhaustion. Furthermore, large-scale events require proactive capacity planning, which involves requesting regional vCPU quota increases well in advance and creating zonal capacity reservations to guarantee compute availability.

Step-by-Step Solution

1
Identify the performance bottleneck
Recognize that CPU utilization is low (~35%) while memory and network sockets are exhausted, indicating an I/O and connection-bound service.
Standard CPU-based autoscaling is ineffective for workloads bound by memory or network socket pool limits.
2
Select the appropriate autoscaling metric
Implement a custom Cloud Monitoring metric tracking open network sockets to trigger MIG autoscaling decisions accurately.
Custom metrics allow autoscaling policies to respond directly to the application's actual resource constraints.
3
Address capacity constraints and quota requirements
Proactively request regional vCPU quota increases from Google Cloud and provision compute capacity reservations prior to stress testing.
Quota increases require lead time for approval, and capacity reservations guarantee VM slot availability in specific zones during large scale-out events.

Key Concept

Custom metric autoscaling for I/O-bound workloads combined with proactive quota management and capacity reservations
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic