Examkin

Soru

Zorluk: OrtaAuto-scaling and Capacity Planning

A media streaming platform hosts a video segment processing service on Compute Engine Managed Instance Groups (MIGs). The application is memory-intensive and network I/O-bound as it fetches uncompressed video segments from Cloud Storage, causing average CPU utilization to stay under 35% even when processing latency increases significantly during live events. Additionally, during sudden traffic spikes, MIG autoscaling attempts fail because requested instance counts exceed the project's regional compute resource limits. Which architectural strategy should a Cloud Architect implement to ensure reliable scaling and adequate capacity during high-demand live events?

  1. Configure the MIG autoscaler using a custom Cloud Monitoring metric based on queue backlog depth, and proactively request regional vCPU quota increases prior to scheduled events.Cevap
  2. B
    Keep the MIG autoscaler configured on target CPU utilization set to a lower threshold of 30%, and purchase 3-year Committed Use Discounts (CUDs) to bypass regional quota restrictions.
  3. C
    Migrate the entire processing application to Google Kubernetes Engine (GKE) with Cluster Autoscaler enabled, relying on default CPU metrics and built-in automatic quota expansions.
  4. D
    Configure the MIG autoscaler using standard HTTP load balancer latency metrics, and rely on regional auto-bursting capabilities to handle capacity shortages during spikes.

Cevap

Configure the MIG autoscaler using a custom Cloud Monitoring metric based on queue backlog depth, and proactively request regional vCPU quota increases prior to scheduled events.
For I/O-bound processing workloads where CPU utilization remains low despite growing backlogs, Compute Engine MIG autoscalers should be configured using custom Cloud Monitoring metrics such as queue depth. Additionally, capacity planning requires requesting regional compute resource quota increases in advance of anticipated peak events to ensure the cloud provider allows instance expansion.

Adım Adım Çözüm

1
Analyze workload resource characteristics and autoscaling metrics.
Identified that the processing service is network I/O and memory bound, meaning CPU utilization does not accurately reflect workload backlog.
Autoscaling I/O-bound workloads based strictly on CPU utilization results in delayed or insufficient scaling.
2
Select an appropriate custom autoscaling metric.
Implemented custom Cloud Monitoring metrics (such as queue backlog depth per instance) to trigger MIG scaling based on actual work waiting to be processed.
Custom metrics directly tied to work queue volume ensure proactive scaling before latency degrades.
3
Evaluate regional capacity and project quota requirements.
Identified that regional vCPU quotas block scaling during peak events.
GCP quotas are hard enforcement limits; regional quota increases must be requested and granted ahead of time to accommodate anticipated capacity needs.

Anahtar Kavram

Custom Metric Autoscaling and Capacity Quota Management
Bu soruyu puanla

Konular

Auto-scaling and Capacity Planning

Bu konudan daha fazla soru