Examkin

Question

Difficulty: MediumCapacity Planning and Infrastructure Workload Scaling Optimization

An enterprise ride-hailing platform ingests real-time driver location updates via an HTTP API hosted on a Google Cloud Compute Engine Managed Instance Group (MIG). During daily rush hours, incoming traffic rapidly increases tenfold, causing severe request latency and dropped connections. Monitoring reveals that VM instance CPU utilization remains around 25%, well below the 70% CPU autoscaling target, because worker threads are blocked waiting for database connection pools. Furthermore, the platform team anticipates an additional fivefold growth in peak load for an upcoming national event. Which workload optimization and capacity planning strategy should you implement to resolve the latency issues and guarantee infrastructure availability?

  1. Reconfigure the MIG autoscaling policy to scale based on Cloud Monitoring custom metrics such as active HTTP requests or connection queue depth, and submit a regional vCPU quota increase request well in advance of the national event.Answer
  2. B
    Retain the current MIG autoscaling policy based strictly on CPU utilization, but reduce the scaling target threshold from 70% down to 20% CPU to force earlier instance scaling during traffic spikes.
  3. C
    Migrate the backend API services to a Google Kubernetes Engine (GKE) cluster managed by Horizontal Pod Autoscaler based on CPU load, relying on cluster autoscaling to automatically bypass regional compute quota limits.
  4. D
    Migrate the application database to Cloud Spanner to eliminate connection bottlenecks, and purchase 3-year Committed Use Discounts (CUDs) sized to match the maximum anticipated peak holiday capacity.

Answer

Reconfigure the MIG autoscaling policy to scale based on custom metrics (such as active HTTP requests or queue depth) rather than CPU utilization, and request a regional vCPU quota increase prior to the peak event.
The correct choice addresses both workload scaling efficiency and capacity planning constraints. For I/O-bound HTTP services, scaling on custom metrics like queue depth or active connection count ensures the MIG expands capacity before latency degrades. Simultaneously, submitting quota requests before anticipated high-demand events prevents automated scaling from hitting regional resource caps.

Step-by-Step Solution

1
Analyze the scaling metric bottleneck.
Identified that CPU utilization is an ineffective scaling metric for I/O-bound workloads where threads wait on database connection pools.
When workloads are constrained by I/O or connection thread pools, CPU usage remains low despite severe request queueing.
2
Select appropriate autoscaling metric for I/O-bound API workloads.
Configured Cloud Monitoring custom metrics reflecting HTTP request rate or queue depth.
Queue depth and active connection counts accurately signal workload pressure for I/O-bound applications, triggering timely instance scaling.
3
Perform proactive capacity planning for anticipated traffic spikes.
Evaluated regional resource quota limits and requested vCPU quota increases prior to the national event.
Autoscaling mechanisms are bounded by GCP regional quotas. If quota increases are not approved prior to high-demand events, instance creation will fail.

Key Concept

Workload-driven autoscaling metric selection and proactive regional quota management for high-burst events.
Estimated Time:1m 30s
Rate this question

Topics

Capacity Planning and Infrastructure Workload Scaling Optimization