Examkin

Question

Difficulty: MediumPlanning Compute Engine Resources and Machine Types

An enterprise is migrating a specialized healthcare analytics platform to Google Cloud Compute Engine. The system architecture requires planning for two distinct compute workloads:

1. Core Data Store: A memory-intensive, stateful relational database running continuously 24/7 with steady, predictable resource requirements for a planned 3-year operational period.
2. Log Processing Pipeline: A stateless, highly fault-tolerant batch job that executes nightly and can resume processing seamlessly if interrupted.

Which TWO compute provisioning decisions should you make to optimize overall costs while maintaining workload availability requirements? (Select 2 answers.)

  1. Select a memory-optimized machine family for the Core Data Store and attach a 3-year Committed Use Discount (CUD) to cover the baseline capacity.Answer
  2. Provision Spot Virtual Machines (VMs) for the batch Log Processing Pipeline instances.Answer
  3. C
    Provision Spot Virtual Machines (VMs) for the Core Data Store instances to minimize baseline compute expenses.
  4. D
    Deploy the Core Data Store on GKE Autopilot worker nodes requiring custom Linux kernel parameters.

Answer

The two correct provisioning choices are selecting a memory-optimized machine family with a 3-year Committed Use Discount (CUD) for the Core Data Store, and using Spot Virtual Machines (VMs) for the Log Processing Pipeline.
Selecting a memory-optimized machine type with a 3-year Committed Use Discount (CUD) accurately addresses the memory-bound, continuous stateful database requirements with maximum cost savings. Provisioning Spot VMs for the batch log processing pipeline leverages massive discounts for stateless, fault-tolerant workloads that easily handle interruptions.

Step-by-Step Solution

1
Analyze the resource and persistence requirements of the Core Data Store.
The database is memory-bound, stateful, and runs continuously 24/7 for 3 years.
Memory-bound stateful applications require memory-optimized VM families to ensure performance, while continuous 24/7 usage over 3 years qualifies for 3-year Committed Use Discounts (CUDs).
2
Analyze the operational requirements of the Log Processing Pipeline.
The batch job is stateless, fault-tolerant, runs periodically, and tolerates unexpected interruptions.
Fault-tolerant batch jobs that can restart or resume seamlessly are prime candidates for Spot VMs, which offer deep discounts in exchange for preemptibility.
3
Evaluate potential anti-patterns for database and cluster management.
Reject Spot VMs for stateful databases due to preemptibility risks, and reject GKE Autopilot when custom kernel parameters are required.
Spot VM preemptions break database availability SLAs, and GKE Autopilot locks down host node OS kernel settings.

Key Concept

Compute Engine Workload Matching and Cost Optimization
Estimated Time:1m 45s
Rate this question

Topics

Planning Compute Engine Resources and Machine Types

More questions from this topic