Question

Difficulty: MediumCapacity Planning and Infrastructure Workload Scaling Optimization

A multinational smart grid utility company ingests smart meter telemetry from millions of households into Google Cloud. During extreme weather events, peak message volumes rapidly surge up to six times baseline within a 15-minute window. The processing layer uses Compute Engine Managed Instance Groups (MIGs) pulling telemetry messages from Pub/Sub queues, and the workload is known to be network I/O-intensive rather than CPU-bound. You need to ensure the architecture scales dynamically and maintains operational capacity without message processing delays during sudden load spikes. Which TWO actions should you recommend?

  1. Configure the MIG autoscaling policy based on Cloud Monitoring metrics measuring Pub/Sub queue depth and unacknowledged message age.Answer
  2. B
    Configure the MIG autoscaling policy strictly around target CPU utilization thresholds set at 80% across compute instances.
  3. Perform proactive quota capacity reviews and submit regional N2 vCPU quota increase requests well in advance of peak weather seasons.Answer
  4. D
    Migrate the processing workers to a multi-region Google Kubernetes Engine cluster with dedicated node pools solely to reduce provisioning overhead.

Answer

The architect should configure MIG autoscaling based on Pub/Sub queue depth and unacknowledged message age metrics, while also submitting proactive regional vCPU quota increase requests in advance of peak seasonal demand.
The correct recommendations focus on aligning the autoscaling trigger metric with the actual bottleneck (Pub/Sub queue depth rather than CPU utilization) and ensuring that cloud infrastructure capacity limits (regional vCPU quotas) are raised prior to predicted traffic surges.

Step-by-Step Solution

1
Analyze workload resource bottlenecks.
Identified that the message consumption workload is network I/O-bound rather than CPU-bound.
I/O-bound workloads do not saturate CPU, making CPU utilization metrics ineffective for autoscaling triggers.
2
Select appropriate autoscaling metric for queue processing.
Choose Pub/Sub queue length and unacknowledged message age as custom Cloud Monitoring scaling signals.
Queue-based metrics directly reflect incoming workload pressure and force the MIG to add instances before backlog breaches SLOs.
3
Verify infrastructure quota headrooms.
Evaluate current regional Compute Engine vCPU quotas against peak scaling targets and request limit increases in advance.
Autoscaling attempts that exceed GCP regional quotas will be denied, causing queue backup regardless of autoscaler settings.

Key Concept

Workload scaling optimization for I/O-bound message consumers requires custom queue depth scaling metrics combined with proactive regional quota management.
Rate this question