Examkin

Question

Difficulty: MediumAuto-scaling and Capacity Planning

A telecommunications company hosts an IoT telemetry ingestion service on a Compute Engine Managed Instance Group (MIG). The processing application is network I/O-bound, where capacity is constrained by open TCP socket connections per virtual machine rather than CPU usage. During scheduled hourly bursts of device reporting, incoming traffic drops connections because auto-scaling fails to trigger in time, and rapid scaling attempts fail due to reaching resource limits. Which TWO architecture modifications should you implement to ensure operational reliability and adequate capacity planning during peak reporting windows?

  1. Configure the MIG autoscaler using a custom Cloud Monitoring metric that tracks active TCP connections per instance.Answer
  2. B
    Set the MIG autoscaler target to scale strictly when average CPU utilization exceeds 85% across all instances.
  3. Submit advance quota increase requests for regional Compute Engine CPU limits to cover maximum anticipated peak capacity.Answer
  4. D
    Re-architect the stateless compute fleet onto a multi-zone GKE cluster using Pod Autoscaling to handle socket management.
  5. E
    Depend on Compute Engine's automated dynamic quota expansion to automatically raise regional limits during peak traffic bursts.

Answer

The correct architecture modifications are to configure the MIG autoscaler using a custom Cloud Monitoring metric tracking active TCP connections per instance, and to submit advance quota increase requests for regional Compute Engine CPU limits.
For I/O-bound workloads constrained by TCP connection limits, configuring custom Cloud Monitoring metrics ensures the autoscaler triggers based on actual socket usage rather than CPU load. Additionally, proactive quota management by requesting regional CPU quota increases in advance guarantees that Compute Engine can provision the necessary instances without hitting quota bounds during peak demand windows.

Step-by-Step Solution

1
Identify the primary bottleneck for the I/O-bound telemetry ingestion service.
Recognized that open TCP connections, not CPU utilization, dictate instance capacity needs.
Default CPU utilization autoscaling is insufficient for network socket-constrained applications.
2
Select the appropriate autoscaling metric.
Configured custom Cloud Monitoring metrics tracking active TCP connections to drive MIG autoscaling.
Custom metrics accurately reflect workload saturation for non-CPU bound services.
3
Perform proactive capacity planning.
Requested regional Compute Engine CPU quota increases prior to peak reporting windows.
Quotas are hard limits evaluated upon instance creation and must be approved in advance to avoid scaling failures.

Key Concept

Auto-scaling network I/O workloads with custom Cloud Monitoring metrics and proactive regional capacity quota planning.
Rate this question

Topics

Auto-scaling and Capacity Planning

More questions from this topic